28 juillet 2025 à 1 août 2025
Fuseau horaire Europe/Paris

Liste des Contributions

264 sur 264 affichés
Exporter en PDF
  1. Francis Bach
    28/07/2025 11:15
    Plenary talk

    Denoising diffusion models have enabled remarkable advances in generative modeling across various domains. These methods rely on a two-step process: first, sampling a noisy version of the data—an easier computational task—and then denoising it, either in a single step or through a sequential procedure. Both stages hinge on the same key component: the score function, which is closely tied to...

    Aller à la page de la contribution
  2. Michael Poss (LIRMM, CNRS)
    28/07/2025 12:15
  3. Victor Spitzer (Université Paris-Saclay (LISN) / Lhyfe)
    28/07/2025 14:00
    Sequential decision making under uncertainty
    Contributed talk

    Lot-Sizing is a class of combinatorial optimization problems encountered in industrial production planning. This work addresses the stochastic capacitated lot-sizing problem with inventory bounds and lost sales. A production problem with uncertain demand is investigated with the assumption that decisions can be regularly updated to adjust to the actual demand being progressively revealed. We...

    Aller à la page de la contribution
  4. Dr Laurent Condat (KAUST)
    28/07/2025 14:00
    Communication-Efficient methods for distributed optimization and federated learning
    Invited talk

    In distributed optimization and machine learning, a large number of machines perform computations in parallel and communicate back and forth with a server. In particular, in federated learning, the distributed training process is run on personal devices such as mobile phones. In this context, communication, that can be slow, costly and unreliable, forms the main bottleneck. To reduce...

    Aller à la page de la contribution
  5. Marc Goerigk (University of Passau)
    28/07/2025 14:00
    Advances in two-stages robust optimization

    Robust and bilevel optimization share the common feature that they involve a certain multilevel structure. Hence, although they model something rather different when used in practice, they seem to have a similar mathematical structure. In this talk, I present some connections between different types of robust problems (static robust problems with and without decision-dependence of their...

    Aller à la page de la contribution
  6. Janosch Ortmann (Université du Québec à Montréal)
    28/07/2025 14:00
    Stochastic Programming
    Contributed talk

    In stochastic programming, scenarios approximate distributions of unknown parameters, but in many applications the number of scenarios required to realistically model the uncertainty makes the optimisation problem numerically intractable. This motivates the scenario reduction problem: by finding a smaller subset of scenarios, reduce the numerical complexity while keeping the error at an...

    Aller à la page de la contribution
  7. Taiji Suzuki (The University of Tokyo / RIKEN AIP)
    28/07/2025 14:00
    Machine learning
    Invited talk
    ML

    In this talk, I will talk about some recent results on distribution optimization methods. First, we will talk about a non-convex optimization technique for a Wasserstein gradient flow (WGF). While WGF is guaranteed to converge to a first-order stationary point, for nonconvex functionals the converged solution does not necessarily satisfy the second-order optimality condition; i.e., it could...

    Aller à la page de la contribution
  8. Prof. Alexandre Jacquillat (MIT)
    28/07/2025 14:00
    Applications in energy, finance or logistics

    This paper optimizes the configuration of large-scale data centers toward cost-effective, reliable and sustainable cloud supply chains. The problem involves placing incoming racks of servers within a data center to maximize demand coverage given space, power and cooling restrictions. We formulate an online integer optimization model to support rack placement decisions. We propose a tractable...

    Aller à la page de la contribution
  9. Wim van Ackooij (EDF Lab Paris-Saclay)
    28/07/2025 14:00
    Chance-constrained programming
    Invited talk

    Several problems in practice are described by a set of controlled state equations. If the problem moreover exhibits uncertainty, one can imagine these state equations to be parametrized by a random event or outcome. One may wish to control the final (random) state and ensure that it hits a desired region of space with large enough probability. Motivated by such a setting, we will discuss the...

    Aller à la page de la contribution
  10. Xinyi Guan (The Hong Kong Polytechnic University)
    28/07/2025 14:00
    Contextual stochastic programming
    Invited talk

    Optimal stopping is the problem of determining when to stop a stochastic system in order to maximize reward, which is of practical importance in domains such as finance, operations management and healthcare. Existing methods for high-dimensional optimal stopping that are popular in practice produce deterministic linear policies -- policies that deterministically stop based on the sign of a...

    Aller à la page de la contribution
  11. Ryan Cory-Wright (Imperial College London)
    28/07/2025 14:30
    Applications in energy, finance or logistics
    Invited talk

    Network design problems involve constructing edges in a transportation or supply chain network to minimize construction and daily operational costs. We study a stochastic version where operational costs are uncertain because of fluctuating demand and estimated as a sample average from historical data. This problem is computationally challenging, and instances with as few as 100 nodes often...

    Aller à la page de la contribution
  12. Austin Stromme (ENSAE Paris)
    28/07/2025 14:30
    Machine learning
    Invited talk
    ML

    The Polyak-Łojasiewicz (PL) constant for a given function exactly characterizes the exponential rate of convergence of gradient flow uniformly over initializations, and has been of major recent interest in optimization and machine learning because it is strictly weaker than strong convexity yet implies many of the same results. In the world of sampling, the log-Sobolev inequality plays an...

    Aller à la page de la contribution
  13. Prof. Omar Mouchtaki (NYU)
    28/07/2025 14:30
    Contextual stochastic programming
    Invited talk

    In this work, we explore a framework for contextual decision-making to study how the relevance and quantity of past data affects the performance of a data-driven policy. We analyze a contextual Newsvendor problem in which a decision-maker needs to trade-off between an underage and an overage cost in the face of uncertain demand. We consider a setting in which past demands observed under "close...

    Aller à la page de la contribution
  14. Pedro Perez Aros (Universidad de Chile and Center for Mathematical Modeling)
    28/07/2025 14:30
    Chance-constrained programming
    Invited talk

    Optimization problems involving uncertainty in the constraints arise in a wide range of applications. A natural framework for handling such uncertainty is through probability functions. However, these functions are often nonsmooth, which poses challenges for both analysis and computation. In this talk, we propose a regularization approach based on the Moreau envelope applied to a scalarization...

    Aller à la page de la contribution
  15. Dr Vassilis M. Charitopoulos (Department of Chemical Engineering, Sargent Centre for Process Systems Engineering, UCL)
    28/07/2025 14:30
    Stochastic Programming
    Contributed talk

    Scenario generation methods constitute an important aspect towards efficient solution of Stochastic Programming (SP) problems and exploitation of big data. The ability of these methods to consistently provide scenario sets which guarantee stability on the solution of the stochastic programs is determinant of their performance. In this context, we present a modification of the existing...

    Aller à la page de la contribution
  16. Lilian Chaves Brandao (CEPEL/UFJF)
    28/07/2025 14:30
    Sequential decision making under uncertainty
    Contributed talk

    Energy planning plays a fundamental role in managing the generation resources in a power system. This planning must meet both present and future energy demands, considering various operational, electrical, environmental, political, and other constraints. The main objective is to allocate resources in a way that minimizes costs while mitigating risks associated with future uncertainties,...

    Aller à la page de la contribution
  17. Dr Ayse Arslan (Inria-Bordeaux/Institut de Mathématiques de Bordeaux)
    28/07/2025 14:45
    Advances in two-stages robust optimization

    In this talk we consider two-stage linear adjustable robust optimization problems with continuous and fixed recourse.
    These problems have been the subject of exact solution approaches, notably, constraint generation (CG) and constraint-and-column generation (CCG).
    Both approaches repose on an exponential-sized reformulation of the problem which uses a large number of constraints or...

    Aller à la page de la contribution
  18. Edouard Oyallon (CNRS, Sorbonne Universite)
    28/07/2025 14:45
    Communication-Efficient methods for distributed optimization and federated learning
    Invited talk

    We consider decentralized optimization over a network of $n$ workers, aiming to minimize
    $$ f(x) = \frac{1}{n} \sum_{i=1}^n f_i(x), $$ where each $f_i$ is $\mu$-strongly convex and $L$-smooth. Most known convergence bounds rely on inner-outer loop structures that alternate synchronized gossip steps with (possibly randomized) local gradient updates, implicitly restricting the algorithm...

    Aller à la page de la contribution
  19. Caroline Geiersbach (University of Hamburg)
    28/07/2025 15:00
    Chance-constrained programming
    Invited talk

    This talk is concerned with a class of risk-neutral stochastic optimization problems defined on a Banach space with almost sure conic-type constraints. This kind of problem appears in the context of optimal control with random differential equation constraints where the state of the system is further constrained almost surely. For this class of problems, we investigate the consistency of...

    Aller à la page de la contribution
  20. Thomas Möllenhoff (RIKEN Center for AI Project)
    28/07/2025 15:00
    Machine learning
    Invited talk
    ML

    ADMM is a popular method for federated deep learning which originated in the 1970s and, even though many new variants of it have been proposed since then, its core algorithmic structure has remained unchanged. In this talk, we present a new way to derive and extend federated ADMM. We propose to use a structure called Bayesian Duality which exploits a duality of the posterior distributions...

    Aller à la page de la contribution
  21. Mathieu Dahan (Georgia Institute of Technology)
    28/07/2025 15:00
    Game theory and equilibrium
    Invited talk

    We study an inspection game of incomplete information, in which an inspector randomizes the allocation of heterogeneous detectors to identify multiple illegal commodities strategically hidden by an adversary within a system (e.g., drugs smuggled in containers). Detectors vary in their detection accuracies, and illegal commodities differ in their associated damage values. The inspector (resp....

    Aller à la page de la contribution
  22. Daniel Mimouni (Mines Paris / IFPEN)
    28/07/2025 15:00
    Stochastic Programming
    Contributed talk

    Scenario tree reduction techniques are essential for achieving a balance between an accurate representation of uncertainties and computational complexity when solving multistage stochastic programming problems. In the realm of available techniques, the Kovacevic and Pichler algorithm (Ann. Oper. Res., 2015) stands out for employing the nested distance, a metric for comparing multistage...

    Aller à la page de la contribution
  23. Prof. Bradley Sturt (University of Illinois Chicago)
    28/07/2025 15:00
    Game theory and equilibrium
    Invited talk

    We consider a pricing problem in which the buyer is strategic: given the seller's pricing policy, the buyer can augment the features that they reveal to the seller in order to obtain a low price for the product. We model the seller's pricing problem as a stochastic program over an infinite-dimensional space of pricing policies in which the radii by which the buyer can strategically perturb...

    Aller à la page de la contribution
  24. Michel De Lara (Ecole des Ponts ParisTech)
    28/07/2025 15:00
    Sequential decision making under uncertainty
    Contributed talk

    In decision problems under incomplete information, actions (identified to payoff vectors indexed by states of nature) and beliefs are naturally paired by bilinear duality. We exploit this duality to analyze the interpersonal comparison of the value of information, using concepts and tools from convex analysis. We characterize each decision-maker (DM) by a closed convex lower set, the...

    Aller à la page de la contribution
  25. Jannis Kurtz (University of Amsterdam)
    28/07/2025 15:10
    Advances in two-stages robust optimization
    Invited talk

    In the realm of robust optimization the k-adaptability approach is one promising method to derive approximate solutions for two-stage robust optimization problems with decisions. Instead of allowing all possible second-stage decisions, the k-adaptability approach aims at calculating a limited set of k such decisions already in the first-stage before the uncertainty reveals. The parameter k can...

    Aller à la page de la contribution
  26. Paul Mangold (Univ. Lille, Inria, CNRS, Centrale Lille, UMR 9189 - CRIStAL, F-59000 Lille, France)
    28/07/2025 15:10
    Communication-Efficient methods for distributed optimization and federated learning
    Invited talk
    • Abstract: In federated learning, multiple users collaboratively train a machine learning model without sharing local data. To reduce communication, users perform multiple local stochastic gradient steps that are then aggregated by a central server. However, due to data heterogeneity, local training introduces bias. In this talk, I will present a novel interpretation of the Federated...
    Aller à la page de la contribution
  27. Joaquim Dias Garcia (Soma Energy)
    28/07/2025 15:30
    Contextual stochastic programming
    Contributed talk

    The Application Driven Learning is a framework that integrates the predictive machine learning model training directly with the decision-making processes, optimizing predictions specifically for the application context.

    We present ApplicationDrivenLearning.jl, a high-performance Julia package that enables efficient experimentation and implementation of the framework, particularly for...

    Aller à la page de la contribution
  28. Dr Piero Visconti (INSA-Rouen)
    28/07/2025 15:30

    We study a class of optimal control problems governed by random semilinear parabolic
    equations with almost sure state constraints in the space of continuous functions. We
    obtain necessary conditions of optimality in the form of a maximum principle with two
    multipliers, one for the state constraint and one for the cost function, the multiplier
    for the state constraint takes values in a...

    Aller à la page de la contribution
  29. Jean Pauphilet (London Business School)
    28/07/2025 15:30
    Applications in energy, finance or logistics
    Invited talk

    Benders Decomposition (BD) is a well-known optimization technique for large-scale two-stage mixed-integer problems by decomposing a problem into a pure integer master problem and a continuous separation problem. To accelerate BD, we propose Random Partial Benders Decomposition (RPBD), a decomposition method that randomly retains a subset of the continuous second-stages variables within the...

    Aller à la page de la contribution
  30. Constantin Philippenko (Inria Paris)
    28/07/2025 15:35
    Communication-Efficient methods for distributed optimization and federated learning
    Invited talk

    We analyze a distributed algorithm to compute a low-rank matrix factorization on $N$ clients, each holding a local dataset $\mathbf{S}^i \in \mathbb{R}^{n_i \times d}$, mathematically, we seek to solve $min_{\mathbf{U}^i \in \mathbb{R}^{n_i\times r}, \mathbf{V}\in \mathbb{R}^{d \times r} } \frac{1}{2} \sum_{i=1}^N \|\mathbf{S}^i - \mathbf{U}^i \mathbf{V}^\top\|^2_{\text{F}}$. Considering a...

    Aller à la page de la contribution
  31. Michael Poss (LIRMM, CNRS)
    28/07/2025 15:35
    Robust Optimization and Machine Learning

    Two-stage robust optimization with integer recourse is a notoriously difficult class of problems, yet modelling many important applications. In this we talk, we discuss how to heuristically solve these problems, solving the adversarial problem and the outer minimization problem through local search algorithms. We focus on the case where all decision variables as well as the uncertainty are...

    Aller à la page de la contribution
  32. Haoyuan Xue (University of Toronto)
    28/07/2025 16:30
    Chance-constrained programming
    Contributed talk

    We consider the wavelength dimensioning problem in wavelength division multiplexing optical networks, which aims to determine the set of wavelengths assigned to each link to accommodate future connection requests under uncertain traffic conditions. To tackle this, we propose a two-stage chance-constrained mixed-integer programming (2S-CCMIP) model that minimizes the total assigned wavelength...

    Aller à la page de la contribution
  33. Peter Barkley (Naval Postgraduate School)
    28/07/2025 16:30
    Stochastic Programming
    Contributed talk

    We apply the recently proposed Coupled Adaptable Backward-Forward-Backward Resolvent Splitting Algorithm (CABRA) to the continuous relaxation of the multi-stage stochastic weapon target assignment problem. Our formulation allows decentralized optimization across weapon platforms with minimal data exchange requirements. The CABRA formulation also allows us to adapt the splitting structure to...

    Aller à la page de la contribution
  34. Bo Zeng (University of Pittsburgh)
    28/07/2025 16:30
    (Distributionally) robust optimization
    Invited talk

    Distributionally robust optimization (DRO) has been recognized as a powerful optimization scheme to handle random factors in decision making. The mainstream solution methods are generally based on duality theory, which might be technically challenging and less intuitive. In this talk, we consider two-stage DRO from the primal perspective, and develop a corresponding decomposition algorithm...

    Aller à la page de la contribution
  35. Mohammad Reza Belbasi (Imperial College Business School)
    28/07/2025 16:30
    Machine learning
    Contributed talk
    ML

    Problem definition: A key challenge in supervised learning is data scarcity, which can cause prediction models to overfit to the training data and perform poorly out of sample. A contemporary approach to combat overfitting is offered by distributionally robust problem formulations that consider all data-generating distributions close to the empirical distribution derived from historical...

    Aller à la page de la contribution
  36. Bartolomeo Stellato (Princeton University)
    28/07/2025 16:30
    Robust Optimization and Machine Learning
    Invited talk

    Robust optimization is a tractable and expressive technique for decision-making under uncertainty, but it can lead to overly conservative decisions when pessimistic assumptions are made on the uncertain parameters. Wasserstein distributionally robust optimization can reduce conservatism by being data-driven, but it often leads to large problems with prohibitive solution times. In this talk, we...

    Aller à la page de la contribution
  37. Dr Lewis Ntaimo (Texas A&M University)
    28/07/2025 16:30
    Applications in energy, finance or logistics
    Invited talk

    Two-stage mean-risk stochastic integer programming (MR-SIP) with endogenous uncertainty involves here-and-now decisions that influence future outcomes and is very challenging to solve. We derive a decomposition method for this class of MR-SIP and apply it to an important problem in wildfire management, namely optimal fuel treatment planning (FTP) under uncertainty. The uncertainty stems from...

    Aller à la page de la contribution
  38. Pedram Farghadani Chaharsooghi (Concordia University)
    28/07/2025 16:30
    Sequential decision making under uncertainty
    Contributed talk

    In this paper, we study the stochastic casualty response planning problem and propose a multi-stage stochastic programming model, where initial decisions—such as the location of alternative care facilities (ACFs) and rescue vehicle assignments—are fixed, while patient assignments and allocations of apheresis machines (AM) for blood extraction are updated dynamically as uncertainty unfolds. In...

    Aller à la page de la contribution
  39. Prof. STEVEN GABRIEL (University of Maryland/NTNU/Aalto University)
    28/07/2025 16:30
    Models and Methods for Stochastic Non-Convex Optimization and Equilibrium Problems
    Invited talk

    We present both the DC algorithm as well as several examples in chance-
    constrained programming as well as other deterministic applications to showcase the
    applicability and success of this new approach. In particular this approach has been successfully applied to solving many deterministic and stochastic mixed complementarity problems (MCPs). MCPs generalize non-cooperative games, the...

    Aller à la page de la contribution
  40. Yiling Zhang
    28/07/2025 17:00
    (Distributionally) robust optimization
    Invited talk

    In this talk, we introduce and study a two-stage distributionally two-stage linear problem with integer recourse, where the objective coefficients are random. The random parameters follow the worst-case distribution belonging to a second-order conic representable ambiguity set of probability distributions. We show that the worst-case recourse objective, under various risk measures, can be...

    Aller à la page de la contribution
  41. Juan-Alberto Estrada-Garcia (University of Michigan)
    28/07/2025 17:00
    Applications in energy, finance or logistics
    Invited talk

    During dry and windy seasons, environmental conditions significantly increase the risk of wildfires, exposing power grids to disruptions caused by transmission line failures. Wildfire propagation exacerbates grid vulnerability, potentially leading to prolonged power outages. To address this challenge, we propose a multi-stage stochastic optimization model that dynamically adjusts transmission...

    Aller à la page de la contribution
  42. Welington de Oliveira (Mines Paris PSL)
    28/07/2025 17:00
    Chance-constrained programming
    Contributed talk

    In this talk, we address joint chance-constrained optimization problems where the only uncertain parameter is the right-hand side coefficients in an inequality system. By leveraging one-dimensional marginals, we construct nonlinear cuts that accurately approximate the probability function, which need not be differentiable or satisfy generalized concavity properties. These cuts are integrated...

    Aller à la page de la contribution
  43. Gregorio Martinez Sempere (Minesparis-psl)
    28/07/2025 17:00
    Machine learning
    ML

    This work explores constrained Wasserstein barycenters (CWB), extending the applicability of the classical Wasserstein barycenter (WB) to pre-required geometric, statistical, or constraints. While the WB problem has been extensively studied, constrained settings pose significant challenges, particularly for nonconvex constraint sets. Building upon the Method of Averaged Marginals (MAM), we...

    Aller à la page de la contribution
  44. Dr Ozlem Cavus (Associate Professor)
    28/07/2025 17:00
    Sequential decision making under uncertainty
    Contributed talk

    Multiarmed bandit problems (MABs) present a class of optimal control problems well-suited for modeling resource allocation under uncertainty. This study explores the application of MABs in the context of clinical trial design. While traditional risk-neutral MABs aim to maximize the expected total number of effective treatments, this study considers the limitations of this objective, as...

    Aller à la page de la contribution
  45. Larry Wigington (Naval Postgraduate School)
    28/07/2025 17:00
    Stochastic Programming
    Invited talk

    We present a computational study exploring methods for solving Stochastic Linear Programs (SLPs) on Graphics Processing Units (GPUs). We examine the operator splitting approach of O’Donoghue et al. and the Primal Dual Hybrid Gradient method of Chambolle and Pock, with the aim of specializing both to exploit the unique sparse structures inherent in SLPs. Our work focuses on adapting these...

    Aller à la page de la contribution
  46. Manish Bansal (Virginia Tech)
    28/07/2025 17:15
    Models and Methods for Stochastic Non-Convex Optimization and Equilibrium Problems
    Invited talk

    We introduce two-stage stochastic min-max and min-min integer programs with bi-parameterized recourse (BTSPs), where the first-stage decisions affect both the objective function and the feasible region of the second-stage problem. To solve these programs efficiently, we introduce Lagrangian-integrated L-shaped ($L^2$) methods, which guarantee exact solutions when the first-stage decisions are...

    Aller à la page de la contribution
  47. M. Julien Grand-Clément (HEC Paris)
    28/07/2025 17:15
    Robust Optimization and Machine Learning
    Invited talk

    In this talk we investigate the tractability of robust Markov Decision Processes (RMDPs) under various structural assumptions on the uncertainty set. Surprisingly, we show that in all generality (i.e. without any assumption on the instantaneous rewards), s-rectangular and sa-rectangular uncertainty sets are the only models of uncertainty that are tractable. Our analysis also shows that...

    Aller à la page de la contribution
  48. Robert Bassett (Naval Postgraduate School)
    28/07/2025 17:30
    Stochastic Programming
    Invited talk

    In this talk we introduce the Matrix Parametrized Proximal Splitting, a proximal splitting method which generalizes a number of splitting methods introduced in the recent literature. Our formulation is capable of constructing convergent splitting algorithms for arbitrary communication constraints imposed between the proximal operators. We review central aspects of the method, state...

    Aller à la page de la contribution
  49. Mme Raneem Madani (Laboratoire Des Signaux Et Systèmes, L2s - Centralesupélec)
    28/07/2025 17:30
    Chance-constrained programming
    Contributed talk

    Optimization problems involving complex variables, when solved, are typically transformed into real variables, often at the expense of convergence rate and interpretability. In this work, we introduce a novel formalism for a prominent problem in stochastic optimization involving complex random variables, which is termed the Complex Chance-Constrained Problem (CCCP). The study specifically...

    Aller à la page de la contribution
  50. Arjun Ramachandra
    28/07/2025 17:30
    (Distributionally) robust optimization
    Invited talk

    Distributionally robust optimization is used to solve decision making problems under uncertainty where the distribution of the uncertain data is itself ambiguous. While several tractable models have been proposed for continuous uncertainty using a convex ambiguity set, fewer results are known for discrete uncertainty. In this work, we identify a class of distributionally robust optimization...

    Aller à la page de la contribution
  51. Yongjia Song (Clemson University)
    28/07/2025 17:30
    Applications in energy, finance or logistics
    Invited talk

    In this talk, we will discuss multi-stage stochastic programming (MSP) models and solution approaches for humanitarian relief logistics planning in hurricane disasters. Specifically, we study how the rolling forecast information can be integrated in an MSP model via the Martingale Model of Forecast Evolution (MMFE) to provide optimal adaptive logistics decision policies. We investigate...

    Aller à la page de la contribution
  52. Vladimir Norkin (V.M.Glushkov Institute of Cybernetics)
    28/07/2025 17:30
    Machine learning
    Contributed talk
    ML

    The study proposes and explores a wide area of application of machine learning methods in applied mathematics, namely, parametric analysis of mathematical models by machine learning methods, in particular, by adaptive kernel support vector machines. The kernel support vector machine is extended by the ability to use a continuum (multivariate parametric) family of kernels to approximate the...

    Aller à la page de la contribution
  53. Raul Garcia (Rice University)
    28/07/2025 17:30
    Sequential decision making under uncertainty
    Contributed talk

    Health organizations (society) prefer to recommend universal screening guidelines to at-risk individuals (patients) for various diseases, with coverage typically provided by third-party payers. However, patients are heterogeneous both in disease risk and disutility associated with screening, resulting in varying health outcomes. Infrequent recommendations leave at-risk patients to decide if to...

    Aller à la page de la contribution
  54. Julien Ancel (Laboratoire de Génie Industriel, Centralesupélec, Université Paris-Sacaly)
    28/07/2025 17:40
    Models and Methods for Stochastic Non-Convex Optimization and Equilibrium Problems
    Contributed talk

    In electricity systems, investment in generation capacity is subject to risk. The distribution of uncertain parameters on which investment decisions depend might not be fully observed in historical values. In Europe, this was recently illustrated by the crisis of exceptionally high power prices during the 2021-2023 period, which was subsequently followed by a regime of extremely low and even...

    Aller à la page de la contribution
  55. Yan Li (Texas A&M University)
    28/07/2025 17:40
    Robust Optimization and Machine Learning
    Invited talk

    The main goal of this talk is to discuss several approaches to formulation of distributionally robust counterparts of Markov decision processes, where the transition kernels are not specified exactly but rather are assumed to be elements of the corresponding ambiguity sets. The intent is to clarify some connections between the game and static formulations of distributionally robust MDPs, and...

    Aller à la page de la contribution
  56. M. Ange Valli (Laboratoire des Signaux et Systèmes (L2S), Université Paris-Saclay, CNRS, CentraleSupélec)
    28/07/2025 18:00
    Chance-constrained programming
    Contributed talk

    In this talk, we present a general method for solving the optimal control problem of trajectory planning for autonomous vehicles in continuous time with robustness to uncertainty. In a precedent work [1], we proposed a formulation as a non-linear optimisation problem with an integral cost function including chance constraints. Our present work uses Pontryagin's maximum principle to solve...

    Aller à la page de la contribution
  57. Jingyi Zhao
    28/07/2025 18:00
    Stochastic integer programming
    Invited talk

    The integration of inventory management and vehicle routing decisions creates a complex combinatorial optimization problem, known as the Inventory Routing Problem (IRP), which is a fundamental challenge in supply chain optimization and has been widely studied over the past decades. However, in the Stochastic IRP (SIRP), where retailer demand varies over time, the problem becomes more...

    Aller à la page de la contribution
  58. Edith Alice Kovács (Budapest University of Technology and Economics)
    28/07/2025 18:00
    Machine learning
    Contributed talk
    ML

    Naive Bayes is one of the most widely used machine learning algorithms, appreciated for its simplicity, efficiency, and ease of interpretation—qualities that make it appealing across various fields. However, Naive Bayes operates under the assumption that the explanatory variables are conditionally independent given the class label, an assumption that often does not hold true in practice.

    To...

    Aller à la page de la contribution
  59. Jingui Xie (Technical University of Munich)
    28/07/2025 18:00
    Sequential decision making under uncertainty
    Contributed talk

    Deciding when to stop medical treatment with uncertain outcomes and predictions is a critical challenge in intensive care units. This research develops a risk-sensitive approach to optimal medical stopping decisions by integrating outcome variability into the decision-making process and incorporating predictive information about the next state. We model the problem using a risk-sensitive...

    Aller à la page de la contribution
  60. Pascal Quach (CentraleSupelec)
    28/07/2025 18:00
    (Distributionally) robust optimization
    Contributed talk

    We propose a decomposition algorithm to approximately solve two-stage distributionally robust optimization problems with mixed-integer ambiguity sets. Such settings are particularly relevant in non-cooperative contexts, such as unplanned disruptions or interdictions, where an adversary reacts to the defender's decisions. DRO is a natural fit, but resulting problems are difficult to reformulate...

    Aller à la page de la contribution
  61. Mathieu Dahan (Georgia Institute of Technology)
    28/07/2025 18:05
    Models and Methods for Stochastic Non-Convex Optimization and Equilibrium Problems
    Invited talk

    We study a two-player discounted zero-sum stochastic game model for dynamic operational planning in military campaigns. At each stage, the players manage multiple commanders who order military actions on objectives that have an open line of control. When a battle over the control of an objective occurs, its stochastic outcome depends on the actions and the enabling support provided by the...

    Aller à la page de la contribution
  62. Aras Selvi (Imperial College London)
    28/07/2025 18:05
    Robust Decision Making in Dynamic Environments
    Invited talk

    Motivated by a housing allocation problem faced by our public-sector partner, the Los Angeles Homeless Services Authority (LAHSA), we study sequential resource allocation in high-stakes social settings, where fairness is critical. Each month, we must allocate limited housing capacity among individuals experiencing homelessness. We model this as a Markov Decision Process (MDP), where...

    Aller à la page de la contribution
  63. Dr Alois Pichler (TU Chemnitz)
    29/07/2025 09:15
    Stochastic Programming
    Plenary talk

    We consider the distance of probability measures from varying angles. We discuss balanced and unbalanced transport, we consider entropic regularization and the maximum mean discrepancy distance.
    Quantization is the approximation of probability measures by simple and discrete measures. The quantization measures behave differently in these metrics – an aspect, which the talk addresses as well.

    Aller à la page de la contribution
  64. Daniel Kuhn (EPFL)
    29/07/2025 10:45
    Taming the Curse of Dimension in Multistage Stochastic Programming

    We introduce contextual stochastic bilevel optimization (CSBO) - a stochastic bilevel optimization framework with the lower-level problem minimizing an expectation conditioned on contextual information and on the upper-level decision variable. We also assume that there may be multiple (or even infinitely many) followers at the lower level. CSBO encapsulates important applications such as...

    Aller à la page de la contribution
  65. Stan URYASEV (Professor at Stony Brook University, USA)
    29/07/2025 10:45
    Stochastic Programming
    Invited talk

    This paper estimates distribution of a response variable conditioned on observing factors (features). We estimate the conditional quantile of the distribution as a mixture (weighted sum) of basis quantile functions with weights depending on factors. The suggested factor model has a closed-form expression. The calibration problem is reduced to conducting quantile regressions for all...

    Aller à la page de la contribution
  66. Darinka Dentcheva (Stevens Institute of Technology)
    29/07/2025 10:45
    Machine learning
    Invited talk
    ML

    Optimization problem with stochastic dominance constraints provide a possibility to shape risk by selecting a benchmark random outcome with a desired distribution. A difficulty arises when no feasible decision results in a distribution that dominates the benchmark. In this talk, we address the problem of choosing a tight relaxation of the stochastic dominance constraint by selecting a feasible...

    Aller à la page de la contribution
  67. Tomás Gutierrez (Lamps Co)
    29/07/2025 10:45
    Applications in energy, finance or logistics

    PolieDRO is a novel analytics framework for classification and regres-
    sion that harnesses the power and flexibility of Data-Driven Distributionally Robust Optimization (DRO) to circumvent the need for regularization hyperparameters. Recent literature shows that traditional machine learning methods such as SVM and (square-root) LASSO can be written as Warserstein-based DRO problems. Inspired...

    Aller à la page de la contribution
  68. Yilin Xue (National University of Singapore)
    29/07/2025 10:45
    (Distributionally) robust optimization
    Invited talk

    This paper shifts focus from the typical approach of maximizing expected reward to minimizing expected regret, aiming to find a solution whose expected reward is close to the oracle. While these two approaches are equivalent when the uncertainty distribution is given, they diverge when accounting for distributional ambiguity, which is characterized by the Wasserstein distance, for enhanced...

    Aller à la page de la contribution
  69. Maria Carolina Bazotte Corgozinho (CIRRELT, Polytechnique Montreal)
    29/07/2025 10:45
    Stochastic Programming
    Contributed talk

    Real-world decision-making problems often involve decision-dependent uncertainty, where the probability distribution of the random vector depends on the model's decisions. Few studies focus on two-stage stochastic programs with this type of endogenous uncertainty, and those that do lack general methodologies. We propose a general method for solving a class of these programs based on random...

    Aller à la page de la contribution
  70. Dr Bismark Singh (University of Southampton)
    29/07/2025 10:45
    Chance-constrained programming
    Contributed talk

    Lagrangian relaxation schemes, coupled with a subgradient procedure, are frequently employed to solve chance-constrained optimization models. Subgradient procedures typically rely on step-size update rules. Although there is extensive research on the properties of these step-size update rules, there is little consensus on which rules are most suitable practically; especially, when the...

    Aller à la page de la contribution
  71. Eduardo Moreno (Google Research)
    29/07/2025 10:45
    Applications in energy, finance or logistics
    Contributed talk

    Stochastic facility location problems with outsourcing costs (SFLPOC) optimize facility placement and customer assignment under demand uncertainty. Excess demand beyond the capacity of a facility incurs outsourcing costs. This work addresses SFLPOC, aiming to minimize overall expected costs (placement, service and outsourcing). We model SFLPOC as a two-stage stochastic program. While prior...

    Aller à la page de la contribution
  72. David Woodruff (UC Davis)
    29/07/2025 10:45
    Stochastic Mixed-Integer Programming
    Invited talk

    MIP-SPPY is software for stochastic programming that can run on a laptop, but uses MPI (message passing interface) to perform well on distributed memory computers. In this talk I will talk about high level changes to the user interface and low level changes to the way we manage interaction between algorithms running asynchronously in parallel.

    Aller à la page de la contribution
  73. Haoming Shen (University of Arkansas, USA)
    29/07/2025 11:09
    Stochastic Programming
    Contributed talk

    AChance constraints yield nonconvex feasible regions in general. In particular, when the uncertain parameters are modeled by a Wasserstein ball, existing studies showed that the distributionally robust (pessimistic) chance constraint admits a mixed-integer conic representation. This talk identifies sufficient conditions that lead to convex feasible regions of chance constraints with...

    Aller à la page de la contribution
  74. Andrzej Ruszczynski (Rutgers University)
    29/07/2025 11:15
    Machine learning
    Invited talk
    ML

    We consider stochastic optimization problems involving an expected value of a nonlinear function of a base random vector and a conditional expectation of another function depending on the base random vector, a dependent random vector, and the decision variables. We call such problems conditional stochastic optimization problems. They arise in many applications, such as uplift modeling,...

    Aller à la page de la contribution
  75. Peyman Mohajerin Esfahani (University of Toronto)
    29/07/2025 11:15
    Machine learning
    Invited talk

    We study a class of learning models known as inverse optimization (IO), where the goal is to replicate the behaviors of a decision-maker (i.e., optimizer) with an unknown objective function. We discuss recent developments in IO concerning convex training losses and optimization algorithms. The main message of this talk is that IO is a rich learning model that can capture complex, potentially...

    Aller à la page de la contribution
  76. Xu Zhou (University College London)
    29/07/2025 11:15
    Applications in energy, finance or logistics
    Contributed talk

    In the context of the global transition towards net-zero emissions, local energy markets (LEMs) offer a practical and effective approach for integrating the increasing penetration of distributed energy resources, such as intermittent renewable generation, energy storage systems, and flexible loads. By facilitating active participation from small-scale consumers, producers, and prosumers in...

    Aller à la page de la contribution
  77. Dr Terry Rockafellar
    29/07/2025 11:15
    Stochastic Programming
    Invited talk

    Least-squares regression is typically formulated as a quadratic program. This talk presents a novel approach for reducing it to a piecewise linear convex minimization problem within the Risk Quadrangle Framework. Evidently, this problem can be reduced to linear programming. Crucially, this is not a heuristic step: the linearized formulation is statistically justified and shown to be equivalent...

    Aller à la page de la contribution
  78. Jean-Paul Watson (Lawrence Livermore National Laboratory)
    29/07/2025 11:15
    Stochastic Mixed-Integer Programming
    Invited talk

    Parallel implementations of scenario-based decomposition strategies are now scalable to thousands and millions of scenarios, thanks to advances in modeling systems (e.g.., Pyomo) and supporting meta-solvers (e.g., mpi-sppy). We describe challenges and their solution considering a large-scale power grid capacity expansion model, where scenarios represent individual days of weather. We discuss...

    Aller à la page de la contribution
  79. Stein-Erik Fleten (Norwegian University of Science and Technology)
    29/07/2025 11:15
    Applications in energy, finance or logistics
    Invited talk

    We report on a two-phase optimization framework for combining short-term hydropower scheduling with offering into the European day-ahead electricity market. We use profiled block bids grouped in exclusive sets. The first phase solves a nonlinear deterministic model that generates a diverse and operationally feasible set of production blocks by accounting for startup costs, opportunity costs,...

    Aller à la page de la contribution
  80. Csaba Fabian (John von Neumann University)
    29/07/2025 11:15
    Chance-constrained programming
    Contributed talk

    Gradient computation of multivariate distribution functions calls for a considerable effort. Hence coordinate descent and derivative-free approaches are attractive. This talk deals with constrained convex problems. We perform random descent steps in an approximation scheme that is an inexact cutting-plane method from a dual viewpoint. We prove that the scheme converges and present a...

    Aller à la page de la contribution
  81. Yifan Hu (EPFL)
    29/07/2025 11:30
    Taming the Curse of Dimension in Multistage Stochastic Programming
    Invited talk

    Algorithms for solving Multistage Stochastic Programming (MSP) are long known to suffer from the curse of dimension and horizon. Stochastic Dual Dynamic Programming (SDDP) for stage-wise independence MSP problem admits a complexity of scales exponentially with respect to the decision dimension $d$ in terms of accuracy $\epsilon$, i.e., $O(\epsilon^{-d})$. On the other hand, Stochastic...

    Aller à la page de la contribution
  82. Mengmeng Li (EPFL)
    29/07/2025 11:33
    Contributed talk

    We study offline reinforcement learning problems with a long-run average reward objective. The state-action pairs generated by any fixed behavioral policy thus follow a Markov chain, and the empirical state-action-next-state distribution satisfies a large deviations principle. We use the rate function of this large deviations principle to construct an uncertainty set for the unknown true...

    Aller à la page de la contribution
  83. Heng Zhang (Paris-Saclay University, CNRS, CentraleSupelec, Laboratory of signals and systems)
    29/07/2025 11:45
    Chance-constrained programming
    Contributed talk

    Chance constraints describe a set of given random inequalities depending on the decision vector satisfied with a large enough probability. They are widely used in decision making under uncertain data in many engineering problems. This talk aims to derive the convexity of chance constraints with elliptically distributed dependent rows via a Gumbel-Hougaard copula. The eventual convexity of...

    Aller à la page de la contribution
  84. Yihang Zhang (University of Southern California)
    29/07/2025 11:45
    Applications in energy, finance or logistics
    Contributed talk

    We propose decomposition algorithms to solve computationally challenging multi-timescale mixed-integer stochastic optimization problems in power system operation, where decisions across different time horizons are coordinated using aggregate state variables. Three distinct decomposition strategies are presented based on the stochastic model: (1) Price-directive decomposition for multi-horizon...

    Aller à la page de la contribution
  85. Bernardo Martin-Iradi (ETH Zurich)
    29/07/2025 11:45
    Applications in energy, finance or logistics
    Invited talk

    Microtransit offers opportunities to enhance urban mobility by combining the reliability of public transit and the flexibility of ride-sharing. This paper optimizes the design and operations of a deviated fixed-route microtransit system that relies on reference lines but can deviate on demand in response to passenger requests. We formulate a Microtransit Network Design (MiND) model via...

    Aller à la page de la contribution
  86. Prof. Joachim Ehrenthal (University of Applied Sciences and Arts Northwestern Switzerland FHNW)
    29/07/2025 11:45
    Stochastic integer programming
    Invited talk

    The Influence Maximization Problem (IMP), which seeks to identify influential nodes in a network to maximize expected information spread, is inherently stochastic and NP-hard—making it a natural candidate for quantum optimization methods. Classical approaches typically formulate IMP as a deterministic optimization problem (e.g., via sample average approximation), then encode it as a Quadratic...

    Aller à la page de la contribution
  87. Milos Kopa
    29/07/2025 11:45
    Applications in energy, finance or logistics
    Contributed talk
    ML

    Stochastic dominance rules are well-characterized and widely used. This work aims to describe and better understand the situations when they do not hold by developing measures of stochastic non-dominance. They quantify the error caused by assuming that one random variable dominates another one when it does not. To calculate them, we search for a hypothetical random variable that satisfies the...

    Aller à la page de la contribution
  88. Prof. Johannes Royset (University of Southern California)
    29/07/2025 11:45
    Stochastic Programming
    Invited talk

    For parameterized mixed-binary optimization problems, we construct local decision rules that prescribe near-optimal courses of action across a set of parameter values. The decision rules stem from solving risk-adaptive training problems over classes of continuous, possibly nonlinear mappings. In asymptotic and nonasymptotic analysis, we establish that the decision rules prescribe near-optimal...

    Aller à la page de la contribution
  89. Angelos Georghiou (University of Cyprus)
    29/07/2025 11:45
    (Distributionally) robust optimization

    The entropic risk measure is commonly used in high-stakes decision-making to account for tail risks. Empirical entropic risk estimator that replaces expectation in the entropic risk measure with sample average underestimates true risk. To correct this bias, a strongly asymptotically consistent bootstrapping procedure is proposed that fits a distribution to the data and then estimates the bias...

    Aller à la page de la contribution
  90. Buse Şen (EPFL)
    29/07/2025 11:55
    Taming the Curse of Dimension in Multistage Stochastic Programming
    Invited talk

    We introduce Multilevel Conditional Compositional Optimization (MCCO) as a new framework for decision-making under uncertainty that combines aspects of multistage stochastic programming and conditional stochastic optimization. MCCO minimizes a nest of conditional expectations and nonlinear cost functions. It finds wide applications in optimal stopping, credit valuation adjustments,...

    Aller à la page de la contribution
  91. Tianyu Wang
    29/07/2025 11:57

    In data-driven optimization, the sample performance of the obtained decision typically incurs an optimistic bias against the true performance, a phenomenon commonly known as the Optimizer's Curse and intimately related to overfitting in machine learning. We develop a general approach that we call Optimizer's Information Criterion (OIC) to correct this bias. OIC generalizes the celebrated...

    Aller à la page de la contribution
  92. Lucas Osmani (Laboratoire des signaux et systèmes)
    29/07/2025 12:15
    Chance-constrained programming
    Contributed talk

    We consider a two-person zero-sum discounted stochastic game with random rewards and known transition probabilities. The players have opposite objectives and are interested in optimizing the expected discounted reward which they can obtain with a given confidence level when both the players play the worst possible move against each other.
    We model such a game problem by defining the...

    Aller à la page de la contribution
  93. Rajmadan Lakshmanan (Chemnitz University of Technology)
    29/07/2025 12:15
    Stochastic Programming
    Contributed talk

    Stochastic dominance is essential in decision-making under uncertainty and quantitative finance, providing a rigorous method for comparing random variables through their distribution functions.
    Despite its importance in decision-making under uncertainty, (higher-order) stochastic dominance is computationally intractable due to infinitely many constraints.
    Our research addresses this by...

    Aller à la page de la contribution
  94. Xian Yu (The Ohio State University)
    29/07/2025 12:21

    We study multistage distributionally robust mixed-integer programs under endogenous uncertainty, where the probability distribution of stage-wise uncertainty depends on the decisions made in previous stages. We first consider two ambiguity sets defined by decision-dependent bounds on the first and second moments of uncertain parameters and by mean and covariance matrix that exactly match...

    Aller à la page de la contribution
  95. 29/07/2025 13:00
  96. Utsav Sadana (Université de Montréal)
    29/07/2025 14:00
    Contextual Stochastic Programming
    Invited talk

    Recent advances in operations research (OR) and machine learning (ML) have spurred interest in integrating prediction algorithms with optimization techniques to address decision-making under uncertainty. This has led to the emergence of contextual optimization, a field focused on data-driven methods that prescribe actions based on the most recently available information. These models appear in...

    Aller à la page de la contribution
  97. Sebastian Maier (University College London)
    29/07/2025 14:00
    Machine learning
    Contributed talk
    ML

    In this paper, we introduce an approach for obtaining probabilistically guaranteed upper and lower bounds on the true optimal value of stopping problems. Bounds of existing simulation-and-regression approaches, such as those based on least squares Monte Carlo and information relaxation, are stochastic in nature and therefore do not come with a finite sample guarantee. Our data-driven approach...

    Aller à la page de la contribution
  98. Bernardo Pagnoncelli (SKEMA Business School)
    29/07/2025 14:00
    Sequential decision making under uncertainty
    Contributed talk

    In this study, we examine multistage problems involving multiple agents, commonly known as stochastic dynamic games. Solving such problems is particularly challenging in real-world scenarios with a large number of interacting agents. We present a general formulation and focus on an incomplete market, heterogeneous agent model with aggregate uncertainty—the Krusell-Smith model. Our numerical...

    Aller à la page de la contribution
  99. Edward Anderson (Imperial College London)
    29/07/2025 14:00
    Stochastic Programming
    Contributed talk

    We consider a situation in which observations are made from an underlying distribution that changes over time. We use a non-parametric model for the changes in distribution and suppose that the change is most likely to involve a small Wasserstein distance between two successive distributions. This leads naturally to a formulation in which we estimate the underlying set of distributions through...

    Aller à la page de la contribution
  100. Jana Junova (Charles University)
    29/07/2025 14:00
    Applications in energy, finance or logistics
    Contributed talk

    We extend portfolio selection models with classical stochastic dominance constraints by allowing a controlled violation of these constraints. This relaxation permits the returns of feasible portfolios to differ from those that stochastically dominate the benchmark within a tolerance measured by the Wasserstein distance. We formulate an optimization problem that incorporates the stochastic...

    Aller à la page de la contribution
  101. Jeff Linderoth (University of Wisconsin-Madison)
    29/07/2025 14:00
    Stochastic integer programming
    Contributed talk

    We consider a two-stage stochastic decision problem where the decision-maker has the opportunity to obtain information about the distribution of the random variables X through a set of discrete actions that we refer to as probing. Specifically, probing allows the decision-maker to observe components of a random vector Y that is jointly-distributed with X. We propose a three-stage optimization...

    Aller à la page de la contribution
  102. LI CHEN
    29/07/2025 14:00
    Applications in energy, finance or logistics
    Invited talk

    Process flexibility has been a well-established supply chain strategy in both theory and practice that enhances responsiveness to demand uncertainty. In this study, we expand the scope of this strategy to supply disruption mitigation by analyzing a long chain system. Specifically, we investigate the effectiveness of long chains in the face of random supply disruptions and demand uncertainty....

    Aller à la page de la contribution
  103. Dr Merve Bodur (University of Edinburgh)
    29/07/2025 14:00
    Stochastic Mixed-Integer Programming
    Invited talk

    Multistage stochastic mixed-integer programs (MSMIPs) can model complex sequential decision-making problems under uncertainty and appear in many applications. However, due to both the stochastic and integer components, their inherent computational challenges require sophisticated solution methodologies. In this talk, we will review recent advances in solving MSMIPs, in particular for the...

    Aller à la page de la contribution
  104. Topias Terho (Aalto University)
    29/07/2025 14:00
    Sequential decision making under uncertainty
    Contributed talk

    An influence diagram is a graph representation of a decision problem that models interdependencies between random events, consequences, and decisions. Recently, two frameworks have been developed to find the optimal decision strategy by transforming the influence diagram into a mixed-integer linear program (MILP). Decision programming (Salo et al., EJOR 299/2, 2022) directly translates the...

    Aller à la page de la contribution
  105. Monika Kaľatová (Charles University, Department of Probability and Mathematical Statistics)
    29/07/2025 14:30
    Applications in energy, finance or logistics
    Contributed talk

    It is a common practice in portfolio optimization to focus on the minimization of losses and risk. However, more advanced models incorporating the second-order stochastic dominance (SSD) constraints have gained increasing attention in last two decades. These constraints identify the portfolios that dominate the benchmark portfolio with respect to SSD. Contrary to that, this paper is focused on...

    Aller à la page de la contribution
  106. Dr Luigi Gallo (Department of Mechanical, Energy, and Management Engineering, University of Calabria, Italy)
    29/07/2025 14:30
    Machine learning
    Contributed talk
    ML

    Decision-makers often face complex problems under uncertainty. Statistical and Machine Learning (ML) tools can support these decisions by reducing the lack of knowledge. However, in many real-world scenarios, the uncertainty itself is induced by the decisions taken. In such cases, standard ML models that provide a priori insights of the problem environment become ineffective. In the context of...

    Aller à la page de la contribution
  107. Anton Kleywegt (Georgia Institute of Technology)
    29/07/2025 14:30
    Stochastic Programming
    Contributed talk

    In derivative-free optimization one has access to a zeroth-order oracle, that is, a black box that takes a feasible point as input and provides the objective value at the point, possibly random, as output, but it provides no derivatives. This setting is encountered in many science and engineering applications, and often each call to the black box is expensive. An important approach to...

    Aller à la page de la contribution
  108. Prof. Napat Rujeerapaiboon (National University of Singapore)
    29/07/2025 14:30
    (Distributionally) robust optimization
    Invited talk

    We study a mechanism design problem where a seller aims to allocate a good to multiple bidders, each with a private value. The seller supports or favors a specific group, referred to as the minority group. Specifically, the seller requires that allocations to the minority group are at least a predetermined fraction (equity level) of those made to the rest of the bidders. Such constraints...

    Aller à la page de la contribution
  109. Benjamin Heymann
    29/07/2025 14:30
    Game theory and equilibrium
    Contributed talk

    When learning to play an imperfect information game, it is often easier to first start with the basic mechanics of the game rules.
    For example, one can play several example rounds with private cards revealed to all players to better understand the basic actions and their effects. Building on this intuition, this paper introduces {\it progressive hiding}, an algorithm that learns to play...

    Aller à la page de la contribution
  110. Paul Malisani (IFP Energies nouvelles)
    29/07/2025 14:30
    Sequential decision making under uncertainty
    Contributed talk

    This paper deals with robust stochastic optimal control problems. The main contribution is an extension of the Progressive Hedging Algorithm (PHA) that enhances out-of-sample robustness while preserving numerical complexity. This extension involves adopting the widespread practice in machine learning of variance penalization for stochastic optimal control problems. Using the Douglas-Rachford...

    Aller à la page de la contribution
  111. Mathis Azéma (CERMICS)
    29/07/2025 14:30
    (Distributionally) robust optimization
    Contributed talk

    The distributionally robust optimization (DRO) framework has emerged as a powerful approach for dealing with uncertainty. In the context of unit commitment, where demand uncertainty affects the right-hand side of constraints, we investigate a DRO approach based on the Wasserstein distance with the $L^2 $-norm. This approach can be addressed using Benders' decomposition as in the risk-neutral...

    Aller à la page de la contribution
  112. Weijun Xie (Georgia Institute of Technology)
    29/07/2025 14:45
    Stochastic Mixed-Integer Programming
    Invited talk

    We study stochastic mixed integer programs with both first-stage and recourse decisions involving mixed integer variables. A new family of Lagrangian cuts, termed “ReLU Lagrangian cuts,” is introduced by reformulating the nonanticipativity constraints using ReLU functions. These cuts can be integrated into scenario decomposition methods. We show that including ReLU Lagrangian cuts is...

    Aller à la page de la contribution
  113. Tianyu Wang
    29/07/2025 14:45
    Contextual Stochastic Programming
    Invited talk

    A fundamental problem in the contextual optimization setting is to evaluate and compare the objective performances of data-driven decision rules. In this regard, we analyze the construction of interval estimates on these performances, as an approach to produce reliable evaluation that handles the underlying statistical uncertainty. Specifically, we systematically compare common plug-in and...

    Aller à la page de la contribution
  114. Yiyin Cao (Xi'an Jiaotong University)
    29/07/2025 15:00
    Game theory and equilibrium
    Contributed talk

    An extension of robust optimization to extensive-form games with payoff uncertainty yields robust extensive-form games. To compute Nash equilibria in behavioral strategies for robust extensive-form games with perfect recall, we acquire from a characterization a polynomial system as a necessary and sufficient condition of Nash equilibrium in robust extensive-form games. As a result of this...

    Aller à la page de la contribution
  115. Guilherme Bodin (PSR)
    29/07/2025 15:00
    Sequential decision making under uncertainty
    Contributed talk

    The Stochastic Dual Dynamic Programming (SDDP) algorithm is widely used to solve multi-stage stochastic problems, such as hydrothermal dispatch in power systems. Due to its iterative nature and the need to handle large volumes of data and multiple future scenarios, SDDP is a computationally intensive method. With the increasing complexity of modern systems and the need to respond to energy...

    Aller à la page de la contribution
  116. Xuan Vinh Doan (The University of Warwick)
    29/07/2025 15:00
    Machine learning
    Contributed talk
    ML

    Federated learning concerns training global models in a decentralized manner. Federated learning is important in many applications, especially when training data come from different sources that cannot be shared with the central server due to restrictions on data sharing. With the increasing capacity of data sources, training samples are usually collected and stored on a regular basis, which...

    Aller à la page de la contribution
  117. Zhengchao Wang (Imperial College London)
    29/07/2025 15:00
    (Distributionally) robust optimization
    Invited talk

    Non-profit organizations play a vital role in addressing global challenges, yet their financial sustainability often hinges on the effectiveness of their fundraising campaigns. We collaborate with a major international non-profit organization to develop and test data-driven approaches to increase the efficiency of their fundraising efforts. Our partner organization conducts multiple annually...

    Aller à la page de la contribution
  118. Franco Quezada (Universidad de Santiago de Chile, ENSTA-Paris)
    29/07/2025 15:00
    Applications in energy, finance or logistics

    We address the Multi-Item Capacitated Lot-Sizing Problem (MCLSP) under decision-dependent uncertainty through a new probing-enhanced stochastic programming framework. In this setting, the demand is strongly correlated with another random vector and the decision-maker can strategically acquire partial information about uncertain demand by selecting component of the correlated random vector to...

    Aller à la page de la contribution
  119. Giorgio Consigli (Khalifa University of Science and Technology)
    29/07/2025 15:00
    Applications in energy, finance or logistics
    Contributed talk

    In an early paper we have studied the correspondence between second order interval stochastic dominance (ISD-2) and interval conditional value-at-risk (ICVaR), a tail risk measure carrying specific properties and generalizing the popular conditional value-at-risk.
    Relying on the ICVaR, in this paper, we present a reinforcement learning approach to solve a trade-off problem based on one side...

    Aller à la page de la contribution
  120. Andrew Schaefer (Rice University)
    29/07/2025 15:00
    Stochastic Programming
    Contributed talk

    A central challenge in multi-stage stochastic programs (MSP) lies in constructing scenario trees that approximate the underlying stochastic process with sufficient accuracy while maintaining computational tractability. In this work, we consider the case where the true stochastic process is discretely distributed, and the topology of the approximating scenario tree is fixed.
    We propose two...

    Aller à la page de la contribution
  121. Haoxiang Yang
    29/07/2025 15:10
    Stochastic Mixed-Integer Programming
    Invited talk

    This work proposes a globally converging cutting-plane algorithm for solving stochastic mixed-integer programs (SMIP) with general mixed-integer state variables. We show that Lagrangian cuts can approximate the convex envelope of the value function. To approximate the nonconvex value function to exactness, we need to iteratively add binary state variables and generate Lagrangian cuts on the...

    Aller à la page de la contribution
  122. Meng Qi (Cornell University)
    29/07/2025 15:10
    Contextual Stochastic Programming
    Invited talk

    This paper revisits the theory of \textit{exact regularization} – where optimal solutions of a regularized convex optimization problem exhibit a phase transition phenomenon and eventually coincide with those of the original unregularized problem (under certain conditions).We examine this phenomenon from a norm-free perspective – instead of adopting norm-related assumptions, our results are...

    Aller à la page de la contribution
  123. Vincent Guigues (FGV)
    29/07/2025 15:30
    Stochastic Programming
    Contributed talk

    In this paper, we introduce a new class of decision rules, referred to as Constant Depth Decision Rules (CDDRs), for multistage optimization under linear constraints with uncertainty-affected right-hand sides. We consider two uncertainty classes: discrete uncertainties which can take at each stage at most a fixed number d of different values, and polytopic uncertainties which, at each stage,...

    Aller à la page de la contribution
  124. M. Ziheng Su (The Chinese University of Hong Kong (CUHK) - Department of Systems Engineering & Engineering Management)
    29/07/2025 15:30
    Game theory and equilibrium
    Contributed talk

    In this paper, we consider a non-collaborative game where each player faces two types of uncertainties: aleatoric uncertainty arising from inherent randomness of underlying data in its own decision-making problem and epistemic uncertainty arising from lack of knowledge and statistical information on the rivals' risk preferences. By assuming that players are risk-averse against aleatoric...

    Aller à la page de la contribution
  125. Cagil Kocyigit (University of Luxembourg)
    29/07/2025 15:30
    (Distributionally) robust optimization
    Invited talk

    We consider the problem of learning, from observational data, a logistic regression model to predict the risk of an adverse outcome under no treatment. These problems arise routinely in public health and the social sciences, e.g., to help prioritize individuals for scarce resources or services. The vast majority of the literature on the topic assumes unconfoundedness, i.e., no unobserved...

    Aller à la page de la contribution
  126. Dawen WU
    29/07/2025 15:30
    Machine learning
    Contributed talk
    ML

    Chance-constrained optimization (CCO) problems are stochastic optimization problems with probabilistic constraints defined by a confidence level $\alpha$. A standard solution approach is to transform the CCO problem into a deterministic optimization problem, which is then solved by a numerical solver. However, this approach becomes computationally expensive when dealing with multiple...

    Aller à la page de la contribution
  127. Morteza Davari
    29/07/2025 15:30
    Sequential decision making under uncertainty
    Contributed talk

    In many real-life situations, such as medical product launches, energy investments, or the rollout of new policies, decision-makers must act before knowing exactly when critical information will become available. We develop new mathematical models that incorporate uncertainty about what will happen and when that uncertainty will resolve. Traditional decision-making tools assume fixed timelines...

    Aller à la page de la contribution
  128. Ward Romeijnders (University of Groningen)
    29/07/2025 15:35
    Stochastic Mixed-Integer Programming
    Invited talk

    We consider multistage stochastic mixed-integer programs. These problems are extremely challenging to solve since the expected cost-to-go functions in these problems are typically non-convex due to the integer decision variables involved. This means that efficient decomposition methods using tools from convex approximations cannot be applied to this problem. For this reason, we construct...

    Aller à la page de la contribution
  129. Huifu Xu (The Chinese University of Hong Kong)
    29/07/2025 16:30
    Plenary talk

    Preference robust optimization (PRO) is a relatively new area of robust optimization. In this talk, I give an overview of recent research on utility-based PRO models and computational methods primarily conducted by my collaborators and myself over the past few years. I begin with a description on one-stage maximin utility PRO model where the true utility function representing the decision...

    Aller à la page de la contribution
  130. 29/07/2025 18:00
  131. Jim Luedtke (University of Wisconsin-Madison)
    30/07/2025 09:15
    Stochastic integer programming
    Plenary talk

    Stochastic integer programs model problems where discrete decisions must be made under uncertainty. This combination provides significant modeling power, leading to wide a wide variety of applications such as supply chain network design, power systems design and operations, and service systems design and operations. This combination also leads to computational challenges due to the need to...

    Aller à la page de la contribution
  132. Bradley Sturt (University of Illinois Chicago)
    30/07/2025 10:45

    Optimal stopping is a fundamental class of stochastic dynamic optimization problems with numerous applications in finance and operations management. We introduce a new approach for solving computationally-demanding stochastic optimal stopping problems with known probability distributions. The approach uses simulation to construct a robust optimization problem that approximates the stochastic...

    Aller à la page de la contribution
  133. Melvyn Sim (NUS Business School)
    30/07/2025 10:45
    Robust Decision Making in Dynamic Environments
    Invited talk

    Over the past two decades, robust optimization techniques have efficiently addressed decision problems under uncertainty, offering high assurance of feasibility without being overly conservative. However, research on estimating parameters for robust decision models from data is relatively scarce. In this paper, we focus on a unified framework for robust decision models that integrate robust...

    Aller à la page de la contribution
  134. Benjamin Heymann
    30/07/2025 10:45
    Sequential decision making under uncertainty
    Contributed talk

    This study explores multi-armed bandit problems under the premise that the decision-maker possesses prior knowledge of the arms' distributions and knows the finite time horizon. These conditions render the problems suitable for stochastic multistage optimization decomposition techniques. On the one hand, multi-armed bandit algorithms are integral to reinforcement learning and are...

    Aller à la page de la contribution
  135. Prof. Phebe Vayanos (University of Southern California)
    30/07/2025 10:45
    Machine learning
    Invited talk
    ML

    Logistic regression models are widely used in the social and behavioral sciences and in high-stakes domains, due to their simplicity and interpretability properties. At the same time, such domains are permeated by distribution shifts, where the distribution generating the data changes between training and deployment. In this paper, we study a distributionally robust logistic regression problem...

    Aller à la page de la contribution
  136. Daniela Bernhard (Friedrich-Alexander Universität Erlangen-Nürnberg)
    30/07/2025 10:45
    (Distributionally) robust optimization
    Contributed talk

    Typically, probability distributions that generate uncertain parameters are uncertain themselves or even unknown. Distributional robustness determines optimized decisions that are protected in a robust fashion against all probability distributions in some appropriately chosen ambiguity set. We consider robust joint chance-constrained optimization problems with discrete probability...

    Aller à la page de la contribution
  137. Prof. Stein W. Wallace (NHH Norwegian School of Economics)
    30/07/2025 10:45
    Modeling flexibility: new developments
    Invited talk

    Much of the stochastic programming literature is on mathematics or algorithms, some is on principal models. But very few are on real use of the models. I discuss the problems that arise when trying to really use stochastic programming, and where I think we are (or should be) headed.

    Aller à la page de la contribution
  138. Lars Hellemo (SINTEF)
    30/07/2025 10:45
    Applications in energy, finance or logistics
    Contributed talk

    The green transition poses many challenges following the introduction of renewable energy sources (RES) and sector coupling (e.g. integrating heat and power). Decision makers need to apply more complex models and to better account for uncertainty, often stemming from the non-dispatchable nature of important RES.

    While solution techniques for stochastic programs are available and demonstated...

    Aller à la page de la contribution
  139. Patrick Combettes (North Carolina State University)
    30/07/2025 10:45
    Stochastic Programming
    Invited talk

    This talk concerns models and convergence principles for dealing with stochasticity in a wide range of algorithms arising in nonlinear analysis and optimization in Hilbert spaces. It proposes a flexible geometric framework within which existing solution methods can be recast and improved, and new ones can be designed. Almost sure weak, strong, and linear convergence results are established in...

    Aller à la page de la contribution
  140. Prof. Martin Gugat (FAU Erlangen)
    30/07/2025 10:45
    Chance-constrained programming
    Invited talk

    Time optimal control is a classical problem in control theory.
    In the case that the initial state is known exactly, the problem is to find a feasible control that steers the system exactly to the prescribed target state as fast as possible. For systems where the initial state is uncertain, the statement of the problem has to be modified to take into account this uncertainty. We replace the...

    Aller à la page de la contribution
  141. Harsha Gangammanavar (Southern Methodist University)
    30/07/2025 11:15
    Sequential decision making under uncertainty
    Contributed talk

    Multistage stochastic linear programming (MSLP) offers a powerful framework for decision-making under uncertainty over time. Sampling-based algorithms provide a practical approach to solving the MSLP problems, particularly in large-scale settings. In this arena, stochastic dual dynamic programming (SDDP) has proven to be very effective. SDDP utilizes randomization to solve a deterministic...

    Aller à la page de la contribution
  142. Holger Heitsch (WIAS Berlin)
    30/07/2025 11:15
    Chance-constrained programming
    Invited talk

    The spherical cap discrepancy is a widely used measure for how uniformly a sample of points on the sphere is distributed. It is particularly important for estimating the integration error for certain classes of functions on the sphere. Being hard to compute, this discrepancy measure is typically replaced by some lower or upper estimates when designing optimal sampling schemes for the uniform...

    Aller à la page de la contribution
  143. Dr Rui Gao (UT Austin)
    30/07/2025 11:15

    Wasserstein distributionally robust optimization (DRO) aims to find robust and generalizable solutions by hedging against data perturbations in Wasserstein distance. Despite its recent empirical success in operations research and machine learning, existing performance guarantees for generic loss functions are either overly conservative because of the curse of dimensionality or plausible only...

    Aller à la page de la contribution
  144. Emanuele Naldi (Università degli studi di Genova)
    30/07/2025 11:15
    Stochastic Programming
    Invited talk

    In this talk, we explore the convergence properties of inexact Jordan-Kinderlehrer-Otto (JKO) schemes and proximal-gradient algorithms in Wasserstein spaces. While the classical JKO scheme assumes exact evaluations at each step, practical implementations rely on approximate solutions due to computational constraints. We analyze two types of inexactness: errors in Wasserstein distance and...

    Aller à la page de la contribution
  145. Daniël Vos (Delft University of Technology)
    30/07/2025 11:15
    Stochastic Programming
    Invited talk
    ML

    In domains such as personalized medicine, historical data is used to learn what treatments to prescribe to maximize positive outcomes. Previous studies have proposed methods for creating prescriptive trees: human-interpretable diagrams that indicate what type of treatment an individual should get based on their measurements. However, a remaining problem is that the models perform worse over...

    Aller à la page de la contribution
  146. Tiago Andrade (PSR)
    30/07/2025 11:15
    Applications in energy, finance or logistics
    Contributed talk

    This study implements and compares single and multicut Benders decomposition (BD) for the generation expansion problem (GEP) involving numerous renewable energy plants, incorporating Time-Varying Dynamic Probabilistic Reserve (DPR). We first identify the conditions under which the second stage (operation problem) of the GEP is convex. However, even when the resulting problem is convex, the...

    Aller à la page de la contribution
  147. Eojin Han (University of Notre Dame)
    30/07/2025 11:15
    (Distributionally) robust optimization
    Contributed talk

    Sequential decision making often requires dynamic policies, which are computationally not tractable in general. Decision rules provide approximate solutions by restricting decisions to simple functions of uncertainties. In this paper, we consider a nonparametric lifting framework where the uncertainty space is lifted to higher dimensions to obtain nonlinear decision rules. Current...

    Aller à la page de la contribution
  148. Alan King (IBM Research)
    30/07/2025 11:30
    Modeling flexibility: new developments
    Invited talk

    Time Series Foundation Models, like ChatGPT for language, are trained on vast amounts of curated time series data and can produce a sequence of high-quality predictions from a single input sequence. Can we apply Time Series Foundation Models to the challenge of generating multistage scenario trees?

    This talk explores a method based on Doob's observation that every discrete-time stochastic...

    Aller à la page de la contribution
  149. Mengmeng Li (EPFL)
    30/07/2025 11:30
    Robust Optimization and Machine Learning
    Invited talk

    We study offline reinforcement learning problems with a long-run average reward objective. The state-action pairs generated by any fixed behavioral policy thus follow a Markov chain, and the empirical state-action-next-state distribution satisfies a large deviations principle. We use the rate function of this large deviations principle to construct an uncertainty set for the unknown true...

    Aller à la page de la contribution
  150. Saverio Salzo (Sapienza Università di Roma)
    30/07/2025 11:45
    Stochastic Programming
    Invited talk

    In this talk, we show novel optimal (or near optimal) convergence rates for a clipped version of the projected stochastic subgradient method. We consider nonsmooth convex problems in Hilbert spaces over possibly unbounded domains, under heavy-tailed noise that possesses only the first $p$ moments for $p \in \left]1,2\right]$. For the last iterate, we establish convergence in expectation with...

    Aller à la page de la contribution
  151. Sergei Ketkov (Department of Business Administration, University of Zurich)
    30/07/2025 11:45
    (Distributionally) robust optimization
    Contributed talk

    We consider a class of stochastic interdiction games between an upper-level decision-maker (referred to as a leader) and a lower-level decision-maker (referred to as a follower), where the follower's objective function coefficients are subject to uncertainty.
    More specifically, unlike traditional deterministic interdiction problem settings, the follower's profits (or costs) in our model...

    Aller à la page de la contribution
  152. Akylas Stratigakos (Imperial College London)
    30/07/2025 11:45
    Applications in energy, finance or logistics
    Contributed talk

    Short-term forecasting models typically assume the availability of input data (features) when they are deployed and in use. However, equipment failures, disruptions, and cyberattacks may lead to missing features when such models are used operationally, which could negatively affect forecast accuracy and result in suboptimal operational decisions. In this paper, we use adaptive robust...

    Aller à la page de la contribution
  153. Nathan Justin (University of Southern California)
    30/07/2025 11:45
    Stochastic Programming
    Invited talk
    ML

    We consider the problem of learning classification trees that are robust to distribution shifts between training and testing/deployment data. This problem arises frequently in high stakes settings such as public health and social work where data is often collected using self-reported surveys which are highly sensitive to e.g., the framing of the questions, the time when and place where the...

    Aller à la page de la contribution
  154. Georg Stadler (Courant Institute, New York University)
    30/07/2025 11:45
    Chance-constrained programming
    Invited talk

    We study optimal control of PDEs under uncertainty with the state variable subject to joint chance constraints. These constraints ensure that the random state variable meets pointwise bounds with high probability. For linear governing PDEs and elliptically distributed random parameters, we prove existence and uniqueness results for almost-everywhere state bounds. We prove variance reduction...

    Aller à la page de la contribution
  155. David Wozabal (Vrije Universiteit Amsterdam)
    30/07/2025 11:45
    Sequential decision making under uncertainty
    Contributed talk

    This paper addresses the computation of tight optimistic bounds for multi-stage stochastic optimization problems using information relaxation duality. We introduce a specific class of penalty functions—bi-linear in decisions and the innovations of the underlying stochastic process—to penalize anticipative policies. Our approach provides a generic framework for deriving such bounds, notably...

    Aller à la page de la contribution
  156. Enza Messina (University of Milano-Bicocca)
    30/07/2025 11:55
    Modeling flexibility: new developments
    Contributed talk

    Flexibility is a critical attribute in solving stochastic programming problems, where decision-making must account for uncertainty and adapt to a wide range of potential future scenarios. Stochastic programming is a practical approach in which decisions are determined prior to the realization of uncertain variables, with subsequent adjustments made through recourse mechanisms once the...

    Aller à la page de la contribution
  157. Enlu Zhou
    30/07/2025 11:55
    Sequential decision making under uncertainty
    Invited talk

    Stochastic optimal control with unknown randomness distributions has been studied for a long time, encompassing robust control, distributionally robust control, and adaptive control. We propose a new episodic Bayesian approach that incorporates Bayesian learning with optimal control. In each episode, the approach learns the randomness distribution with a Bayesian posterior and subsequently...

    Aller à la page de la contribution
  158. Axel Friedrich Wolter (University of Konstanz)
    30/07/2025 12:15
    Sequential decision making under uncertainty

    We study reinforcement learning by combining recent advances in regularized linear programming formulations with the classical theory of stochastic approximation.
    Motivated by the challenge of designing algorithms that leverage off-policy data while maintaining on-policy exploration, we propose PGDA-RL, a novel primal-dual Projected Gradient Descent-Ascent algorithm for solving regularized...

    Aller à la page de la contribution
  159. Minh Bui (Institut für Mathematik und Wissenschaftliches Rechnen, Universität Graz)
    30/07/2025 12:15
    Stochastic Programming
    Invited talk

    We discuss an approach for designing block-activated algorithms for solving the monotone multi-stage stochastic variational inequalities in extensive form proposed by Rockafellar and Wets. Advantages over the classical progressive hedging algorithm will be discussed.

    Aller à la page de la contribution
  160. Monika Matoušková
    30/07/2025 12:15
    (Distributionally) robust optimization
    Contributed talk

    In logistics and transportation, scheduling tasks with fixed start times is a common challenge, often complicated by unpredictable delays. To tackle these issues, we need robust optimization techniques that can adapt to real-world uncertainties.
    Initially, we explore an operational FIS problem where job completion times are influenced by random delays, modeled using Archimedean copulas to...

    Aller à la page de la contribution
  161. Rui Chen (Chinese University of Hong Kong, Shenzhen)
    30/07/2025 12:15
    Machine learning
    Invited talk
    ML

    Causal structure learning, the task of inferring causal relationships from data, is computationally challenging due to its combinatorial nature. State-of-the-art integer programming formulations suffer from exponential growth in the number of variables and constraints, while traditional column generation approaches struggle with the complexity of solving mixed-integer nonlinear programming...

    Aller à la page de la contribution
  162. René Henrion
    30/07/2025 12:15
    Chance-constrained programming
    Invited talk

    Sweeping processe have been introduced by J.J. Moreau in 1971. These are special differential inclusions where the set-valued right-hand side is represented by the normal cone to some moving set. We consider the optimal control of polyhedral sweeping processe subject to a terminal state constraint. We shall assume that the control is affected by a random perturbation so that the terminal state...

    Aller à la page de la contribution
  163. Gülin Yurter (University of Twente)
    30/07/2025 12:15
    Applications in energy, finance or logistics
    Contributed talk

    In recent years, many companies have committed to renewable energy procurement targets, which usually require a certain fraction of the annual demand to be met by renewables. This annual approach overlooks the temporal fluctuations in energy supply and demand, leading to a growing interest in 24/7 targets that aim to match every consumed kilo-watt hour with carbon-free electricity sources at...

    Aller à la page de la contribution
  164. Irina Wang (Princeton University)
    30/07/2025 12:20
    Robust Decision Making in Dynamic Environments

    We present a data-driven technique to automatically learn uncertainty sets in dynamic decision making under uncertainty. We formulate the learning problem as a control design problem where the control policy involves solving a robust optimization problem parametrized by the past disturbances, as well as the parameters of the uncertainty set. We propose a learning procedure to dynamically...

    Aller à la page de la contribution
  165. Francesca Maggioni (Department of Management, Information and Production Engineering, University of Bergamo)
    31/07/2025 09:15
    Stochastic Programming
    Invited talk

    Many real world decision problems are dynamic and affected by uncertainty. Stochastic Programming provides a powerful approach to handle this uncertainty within a multi-period decision framework. However, as the number of stages increases, the computational complexity of these problems grows exponentially, posing significant challenges. To tackle this, approximation techniques are often used...

    Aller à la page de la contribution
  166. Titus Pinta (ENSTA)
    31/07/2025 10:45
    Stochastic Programming
    Contributed talk

    We introduce a new framework for analyzing (Quasi-}Newton type methods applied to non-smooth optimization problems. The source of randomness comes from the evaluation of the (approximation) of the Hessian. We derive, using a variant of Chernoff bounds for stopping times, expectation and probability bounds for the random variable representing the number of iterations of the algorithm until...

    Aller à la page de la contribution
  167. Rodrigo Maulen Soto
    31/07/2025 10:45
    Stochastic Programming
    Contributed talk

    We analyze the global and local behavior of gradient-like flows under stochastic errors towards the aim of solving convex optimization problems with noisy gradient input. We first study the unconstrained differentiable convex case, using a stochastic differential equation where the drift term is minus the gradient of the objective function and the diffusion term is either bounded or...

    Aller à la page de la contribution
  168. Tito Homem-de-Mello (Universidad Adolfo Ibáñez)
    31/07/2025 10:45
    Contextual Stochastic Programming
    Invited talk

    We consider the class of two-stage stochastic programs with uncertainty only on the right-hand side. Such a class encompasses practical many problems, especially in inventory models. We show that, under certain conditions, there exist an optimal scenario, in the sense that solving the problem with that scenario yields the same optimal solution as the original problem. In the case data-driven...

    Aller à la page de la contribution
  169. François Pacaud (Mines Paris - PSL)
    31/07/2025 10:45
    Applications in energy, finance or logistics
    Contributed talk

    We apply computational techniques of convex stochastic optimization to optimal operation and valuation of electricity storages in the face of uncertain electricity prices. Our approach is applicable to various specifications of storages, and it allows for e.g. hard constraints on storage capacity and charging speed. Our valuations are based on the indifference pricing principle, which builds...

    Aller à la page de la contribution
  170. Dmitrii Golembiovskii (Moscow State University named Lomonosov)
    31/07/2025 10:45
    Sequential decision making under uncertainty
    Contributed talk

    An implied distribution of the underlying asset price for the options expiration moment can be obtained from the market option prices [1]. However, exchange-traded options rarely expire more often than once a month. It is not enough for planning dynamic decisions in many cases. In [2] implied calibration of the dynamic ARMA(1,1)-GARCH(1,1) model using market prices of options of different...

    Aller à la page de la contribution
  171. Ya-Ping Hsieh (ETH Zürich)
    31/07/2025 10:45
    Machine learning
    Invited talk
    ML

    The score function plays a central role in modern generative modeling, particularly in diffusion models and related score-based methods. Despite its theoretical appeal, learning the score function in practice is notoriously difficult: it is sensitive to hyperparameter choices and prone to various forms of instability that often require ad hoc corrections.

    In this talk, we study score...

    Aller à la page de la contribution
  172. JIANQIANG CHENG (University of Arizona)
    31/07/2025 10:45
    Computationally Efficient Approaches for Distributionally Robust Optimization

    This talk focus on developing computationally efficient approximations for solving optimization problem under uncertainty. We first present a novel and simple modeling method called harmonizing optimization (HO), which integrates SAA and DRO with moment by adaptively adjusting the weights of data and information based on sample size N. This allows HO to amplify data effects in large samples...

    Aller à la page de la contribution
  173. Giovanni Pantuso (University of Copenhagen)
    31/07/2025 10:45
    Stochastic Optimization under Decision-Dependent Uncertainty
    Invited talk

    In this talk we provide an introduction to stochastic optimization problems with decision-dependent uncertainty.
    We review the main lines of research in terms of both methodology and applications.
    Our discussion will center on a taxonomy of these problems, distinguishing between uncertainties that affect probabilities (Type 1) and those that influence resolution time (Type 2)....

    Aller à la page de la contribution
  174. Mike Hewitt (Loyola University Chicago)
    31/07/2025 10:45
    Decomposition methods for solving Stochastic Programming problems in Logistics and Transportation
    Invited talk

    Freight transportation is one of the critical enablers of trade, both global and domestic. Given that, it regularly constitutes significant portions of the gross domestic product of countries. Due to its large scale, small percentage improvements in the efficiency of freight transportation operations can lead to large monetary savings as well as reduced environmental impacts. At the same time,...

    Aller à la page de la contribution
  175. Young-Heon Kim (University of British Columbia)
    31/07/2025 11:15
    Machine learning
    Invited talk
    ML

    The supercooled Stefan problem describes freezing of supercooled water. In contrast to the Stefan problem that describes melting of ice, the supercooled problem exhibits unstable behaviour which makes the usual PDE methods break down. We will discuss some recent progress which employs a stochastic version of optimal transport, involving optimal stopping times of the Brownian motion.

    Aller à la page de la contribution
  176. Nihat Oner (University of Luxembourg)
    31/07/2025 11:15
    Sequential decision making under uncertainty
    Contributed talk

    Cancer is the second leading cause of death in the world. Unfortunately, the projections from the International Agency for Research on Cancer (IARC) indicate a rising trend for new cancer cases in the following years. Among the various cancer treatments, chemotherapy is one of the most effective treatments for numerous cancer types. It generally contains one or more prescribed molecules...

    Aller à la page de la contribution
  177. Grani Adiwena Hanasusanto (University of Illinois Urbana-Champaign)
    31/07/2025 11:15
    Contextual Stochastic Programming
    Invited talk

    We study contextual stochastic optimization problems in which the joint distribution of uncertain parameters and side information covariates is modeled as a mixture of Gaussians. In a data-driven setting, the parameters of this distribution are unknown and must be estimated from historical data. To mitigate the adverse effects of estimation errors and improve out-of-sample performance, we...

    Aller à la page de la contribution
  178. João Marcelo Leal Gomes Leite (Universidade de São Paulo (USP))
    31/07/2025 11:15
    Applications in energy, finance or logistics
    Contributed talk

    Decision support models are essential for assessing logistics infrastructure projects, such as intermodal terminals. This study proposes a methodology to analyze the multiple impacts of strategic decisions on terminal location in conjunction with the tactical problem of designing an intermodal service. The case study of Mato Grosso’s soybean export logistics network was structured as a Markov...

    Aller à la page de la contribution
  179. Jonathan Hornewall (École Des Ponts)
    31/07/2025 11:15
    Stochastic Programming

    We present a novel method for sampling the optimal solution of unconstrained, strictly convex optimization problems with random parameters. The motivating application are methods in two-
    stage stochastic programming, which often rely on computing (the expectation of) optimal dual variables for linear programs with random coefficients.

    Conventional methods typically proceed by generating...

    Aller à la page de la contribution
  180. Bernardo Freitas Paulo da Costa (Fundação Getulio Vargas)
    31/07/2025 11:15
    Stochastic Programming
    Contributed talk

    The use of linear decision rules is an attractive alternative to multistage decision making under uncertainty, combining simplicity and interpretability of the policies and computational tractability. We introduce a modeling extension to the LinearDecisionRules.jl package that allows the user to formulate and optimize value functions in this framework. The extension also simplifies...

    Aller à la page de la contribution
  181. Miguel Lejeune (George Washington University)
    31/07/2025 11:30
    Stochastic Optimization under Decision-Dependent Uncertainty
    Invited talk

    We study distributionally robust chance-constrained optimization problems under decision-dependent uncertainty. The focus is on decision-dependent uncertainty of Type 1 in which the support of the random variables is decision-dependent. We study various types of coupling functions and ambiguity. We derive computationally tractable equivalent reformulations and design customized solution...

    Aller à la page de la contribution
  182. Daniele Manerba (Università degli Studi di Brescia)
    31/07/2025 11:30
    Decomposition methods for solving Stochastic Programming problems in Logistics and Transportation
    Invited talk

    A consolidated business model for managing e-commerce logistics involves the combination of forward-and-reverse operations, where the collection of returns is ensured along with the distribution of products, and the use of hub-and-spoke networks, in which both distribution and collection demand from many customers are aggregated into intermediate hubs. In this context, we study a complex...

    Aller à la page de la contribution
  183. Ruiwei Jiang (University of Michigan)
    31/07/2025 11:30
    Computationally Efficient Approaches for Distributionally Robust Optimization
    Invited talk

    We revisit the value of stochastic solutions (VSS) in adaptive stochastic optimization. Given a fixed decision, VSS evaluates its suboptimality in contrast to an optimal solution with knowledge of the underlying probability distribution. For example, for a decision given by the sample average approximation (SAA), VSS interprets the value of collecting more data for better decision-making. When...

    Aller à la page de la contribution
  184. Chuangyin Dang (City University of Hong Kong)
    31/07/2025 11:45
    Stochastic Programming
    Contributed talk

    The linear tracing procedure plays an essential role in Harsanyi and Selten's (1988) equilibrium selection theory. The concept of proper equilibrium was formulated by Myerson (1978), which is able to eliminate some counterintuitive perfect equilibria in normal-form games. An extension of proper equilibrium to stochastic games leads to a notion of proper Markov perfect equilibrium. To select...

    Aller à la page de la contribution
  185. Camila Martinez Parra (RTE-France)
    31/07/2025 11:45
    Sequential decision making under uncertainty
    Contributed talk

    The increasing penetration of renewable energy sources in power systems amplifies the need for storage to manage their inherent intermittency. In this context, evaluating the opportunity cost of stored energy—commonly referred to as usage values—becomes essential. These values can be computed by solving a multistage stochastic optimization problem, where uncertainty arises from net demand (the...

    Aller à la page de la contribution
  186. Guzin Bayraksan (The Ohio State University)
    31/07/2025 11:45
    Contextual Stochastic Programming
    Invited talk

    We consider data-driven decision-making that incorporates a prediction model within the 1-Wasserstein distributionally robust optimization (DRO) given joint observations of uncertain parameters and covariates using regression residuals in a streaming-data setting. In this setting, additional data becomes available and allows decisions to adapt to the growing knowledge of the underlying...

    Aller à la page de la contribution
  187. Jia-Jie Zhu (Weierstrass Institute, Berlin)
    31/07/2025 11:45
    Machine learning
    Invited talk
    ML

    Many problems in machine learning can be framed as variational problems that minimize the relative entropy between two probability measures. Many recent works have exploited the connection between the (Otto-)Wasserstein gradient flow of the Kullback-–Leibler (KL) divergence and various sampling, Bayesian inference, and generative modeling algorithms. In this talk, I will first contrast the...

    Aller à la page de la contribution
  188. M. Santo Saraceno (University of Brescia)
    31/07/2025 11:45
    Applications in energy, finance or logistics
    Contributed talk

    This work presents a chance-constraint model for the management of Energy
    Communities, focusing on prosumers and peer-to-peer electricity exchanges.

    The model aims to minimize the total operation costs of the community, while ensuring
    energy balance and satisfying technical constraints related to local production and the
    energy exchanges both inside the community and with the main...

    Aller à la page de la contribution
  189. Dr Liviu Aolaritei (UC Berkeley)
    31/07/2025 11:45
    Stochastic Programming
    Contributed talk

    Importance Sampling (IS) is a widely used variance reduction technique for enhancing the efficiency of Monte Carlo methods, particularly in rare-event simulation and related applications. Despite its power, the performance of IS is often highly sensitive to the choice of the proposal distribution and frequently requires stochastic calibration techniques. While the design and analysis of IS...

    Aller à la page de la contribution
  190. Paolo Beatrici (University of Bergamo)
    31/07/2025 11:55
    Decomposition methods for solving Stochastic Programming problems in Logistics and Transportation
    Invited talk

    In this study, we examine the optimization of fleet size and mix, together with vehicle routing, under uncertain demand conditions, with explicit consideration of sustainability aspects in the context of Last Mile logistics. We propose a two-stage bi-objective stochastic mixed-integer programming model that simultaneously minimizes total costs and vehicle emissions associated with delivery...

    Aller à la page de la contribution
  191. Beste Basciftci (University of Iowa)
    31/07/2025 11:55
    Computationally Efficient Approaches for Distributionally Robust Optimization
    Invited talk

    We consider a two-stage distributionally robust optimization (DRO) model with multimodal uncertainty, where both the mode probabilities and uncertainty distributions could be affected by the first-stage decisions. To address this setting, we propose a generic framework by introducing a ϕ-divergence based ambiguity set to characterize the decision-dependent mode probabilities and further...

    Aller à la page de la contribution
  192. Xian Yu (The Ohio State University)
    31/07/2025 11:55
    Stochastic Optimization under Decision-Dependent Uncertainty
    Invited talk

    We consider a residuals-based distributionally robust optimization model, where the underlying uncertainty depends on both covariate information and our decisions. We adopt regression models to learn the latent decision dependency and construct a nominal distribution (thereby ambiguity sets) around the learned model using empirical residuals from the regressions. Ambiguity sets can be formed...

    Aller à la page de la contribution
  193. André Diniz (CEPEL/UERJ)
    31/07/2025 12:15
    Applications in energy, finance or logistics
    Contributed talk

    The increasing penetration of wind and solar generation amplifies the uncertainty in the short term unit commitment problem for hydrothermal power systems, challenging, for example, schedules that are given by deterministic approaches, as is the case of the day ahead-unit commitment model (DESSEM) used for the official dispatch and price setting of the Brazilian system, over a seven day...

    Aller à la page de la contribution
  194. Irene Aldridge (Cornell University, Cambridge University)
    31/07/2025 12:15
    Contextual stochastic programming
    Contributed talk

    This paper proposes an Optimize-then-Predict framework in which we identify the optimal decision before predicting or observing the realized values. The optimization part can be run in low-demand environments, saving computational time during runtime. We also propose computationally efficient inferences for the evaluation of model performance.

    This paper shows that in any optimization...

    Aller à la page de la contribution
  195. Hoda Bidkhori (George Mason University)
    31/07/2025 12:15
    Sequential decision making under uncertainty
    Contributed talk

    Exactly and asymptotically optimal algorithms are developed for robust detection of changes in non-stationary processes. In non-stationary processes, the distribution of the data after change varies with time. The decision maker does not have access to precise information on the post-change distribution. It is shown that if the post-change non-stationary family has a distribution that is least...

    Aller à la page de la contribution
  196. Rosario Paradiso (Vrije Universiteit Amsterdam)
    31/07/2025 12:20
    Stochastic Optimization under Decision-Dependent Uncertainty
    Contributed talk

    In uncertain optimization problems with decision-dependent information discovery, the decision-maker can influence when information is revealed, unlike the classic setting where uncertain parameters are revealed according to a prescribed filtration. This work examines two-stage robust optimization problems with decision-dependent information discovery, focusing on uncertainty in the objective...

    Aller à la page de la contribution
  197. Jiali Deng
    31/07/2025 12:20
    Decomposition methods for solving Stochastic Programming problems in Logistics and Transportation
    Invited talk

    The problem of pricing mobility services has attracted significant attention. In most studies, uncertain demand is modeled as an exogenous random variable with known distribution. This assumption overlooks the likely effect of prices on user adoption decisions. To address this dependency, we formulate the pricing problem as a stochastic program with decision-dependent demand uncertainty....

    Aller à la page de la contribution
  198. 31/07/2025 13:00
  199. Erick Delage (HEC Montréal)
    01/08/2025 09:15
    Stochastic Programming
    Plenary talk

    This talk surveys recent developments in reinforcement learning (RL) methods for risk-aware model-free decision-making in Markov decision processes (MDPs). In the discounted setting, we adapt two popular risk neutral RL methods to account for risk aversion. The first approach minimizes a dynamic utility-based shortfall risk measure, while the other optimizes a specific quantile of the total...

    Aller à la page de la contribution
  200. Dr Maximilian Schiffer (TUM)
    01/08/2025 10:45
    Invited talk

    In this talk, we will bridge the gap between combinatorial optimization and machine learning to derive policies for contextual multi-stage decision-making problems that arise in various stochastic settings, including transportation, control, and supply chain management. We will discuss how to encode effective policies by embedding combinatorial optimization layers into neural networks and...

    Aller à la page de la contribution
  201. M. Renjie Chen (The Chinese University of Hong Kong)
    01/08/2025 10:45
    (Distributionally) robust optimization
    Contributed talk

    We propose a distributionally robust formulation for the simultaneous estimation of the covariance and precision matrix of a random vector. The proposed model minimizes the worst-case weighted sum of the Stein's loss of the precision matrix estimator and the Frobenius loss of the covariance estimator against all distributions from an ambiguity set centered at the empirical distribution. The...

    Aller à la page de la contribution
  202. Hanbin Yang (The Chinese University of Hong Kong, Shenzhen)
    01/08/2025 10:45
    Stochastic integer programming
    Invited talk

    Multistage stochastic mixed-integer programming (MS-MIP) is a powerful framework for sequential decision-making under uncertainty, yet its computational complexity poses significant challenges. This paper presents a convergent cutting-plane algorithm for solving MS-MIP with binary state variables. Our method exploits a geometric interpretation of the convex envelope of value functions and the...

    Aller à la page de la contribution
  203. Théo Molfessis (École Polytechnique)
    01/08/2025 10:45
    Stochastic Programming
    Contributed talk

    Progressive Hedging and Proximal Decomposition are popular splitting methods for large-scale stochastic optimization. We present a formal equivalence between Progressive Hedging and Proximal Decomposition when the nonanticipativity constraint is a subspace, as well as a result of linear convergence of their bundle versions under standard error-bound assumptions once an infinite null-step tail...

    Aller à la page de la contribution
  204. Peter Schütz (Norwegian University of Science and Technology)
    01/08/2025 10:45
    Applications in energy, finance or logistics
    Contributed talk

    We study the problem of locating charging stations for battery-electric heavy-duty vehicles (BEHDV) under uncertainty in both demand and available power grid capacity. The problem can be formulated as a two-stage stochastic problem where the first-stage decision is to determine the future locations of the charging stations. After the locations have been determined, information about demand and...

    Aller à la page de la contribution
  205. Prof. Asgeir Tomasgard (NTNU)
    01/08/2025 10:45
    Multihorizon Stochastic Programming: Models, Algorithms, and Applications
    Invited talk

    This presentation addresses stochastic optimization models for the energy transition, focusing on multiscale multihorizon systems. We model both long-term investment decisions—such as renewable generation, carbon capture, and decarbonized transport—and short-term operations like storage management and system balancing. To handle uncertainty across both scales, we introduce multihorizon...

    Aller à la page de la contribution
  206. Dr Philip Thompson (FGV-EMAp)
    01/08/2025 10:45
    Machine learning
    Contributed talk
    ML

    The notions of relative (weak) convexity and variation (e.g., Lipschitzness and smoothness) have been successfully applied to some optimization problems, including applications in machine learning. While typically harder to prove, these properties encode better dependence of the objective with respect to the intrinsic geometry of the problem. We review previous analysis of the mirror descent...

    Aller à la page de la contribution
  207. Diego Jiménez (SKEMA Business School - KU Leuven)
    01/08/2025 10:45
    Contextual stochastic programming
    Contributed talk

    The recent interest in contextual optimization problems, where randomness is associated with side information, has led to two primary strategies for formulation and solution. The first, estimate-then-optimize, separates the estimation of the problem's parameters from the optimization process. The second, decision-focused optimization, integrates the optimization problem's structure directly...

    Aller à la page de la contribution
  208. Dr Christian Füllner (Karlsruhe Institute of Technology)
    01/08/2025 10:45
    Sequential decision making under uncertainty
    Invited talk

    We consider the generation of cuts in stochastic dual dynamic programming (SDDP) for multistage stochastic linear programming problems with stagewise dependent uncertainty in the right-hand side described by log-linear (or geometric) autoregressive processes. We show that it is possible to develop tractable closed-form cut formulas in this case. The cuts are linear in all decision variables,...

    Aller à la page de la contribution
  209. Kibaek Kim (Argonne National Laboratory)
    01/08/2025 11:15
    Stochastic integer programming
    Invited talk

    Large-scale mixed-integer programs (MIPs) pose significant computational challenges due to their complexity, nonconvexity, and nonsmooth dual functions. Traditional decomposition methods often suffer from slow convergence, sensitivity to parameters, and instability. We propose Bundle-ADMM, a novel method combining the Alternating Direction Method of Multipliers (ADMM) and bundle techniques to...

    Aller à la page de la contribution
  210. Prof. Jonathan Yu-Meng Li (Telfer School of Management, University of Ottawa)
    01/08/2025 11:15
    Contextual stochastic programming
    Contributed talk

    We present a unified framework and practical solution for conditional risk minimization with side information—interpretable, tractable, and backed by finite-sample statistical guarantees.

    Aller à la page de la contribution
  211. Dr Claudia Sagastizábal (IMECC)
    01/08/2025 11:15
    Stochastic Programming
    Contributed talk

    Many popular splitting methods for large-scale stochastic optimization are derived from Spingarn's partial inverse framework. Well-known and popular methods such as Progressive Hedging and Proximal Decomposition are paradigmatic examples of this class. We present lessons learned by examining Spingarn's framework from a dual perspective, inspired from bundle methods in nonsmooth...

    Aller à la page de la contribution
  212. Aleksandr Lobanov (MIPT)
    01/08/2025 11:15
    Machine learning
    Contributed talk
    ML

    The gradient descent (GD) method -- is a fundamental and likely the most popular optimization algorithm in machine learning (ML), with a history traced back to a paper in 1847 (Cauchy, 1847). In this paper, we provide an improved convergence analysis of gradient descent and its variants, assuming generalized smoothness (L0,L1). In particular, we show that GD has the following behavior of...

    Aller à la page de la contribution
  213. M. Gabriel Vidigal (PSR)
    01/08/2025 11:15
    Sequential decision making under uncertainty
    Invited talk

    Hydropower scheduling is an important application of stochastic dynamic programming, involving large optimization problems with intertemporal dependency and complex dynamics. The nature of this problem is naturally seasonal (and therefore cyclic), and even though most real-world use cases usually apply discretization strategies (such as monthly, weekly or daily time steps) it is reasonable to...

    Aller à la page de la contribution
  214. Florian Vincent (Inria Grenoble, Laboratoire Jean Kuntzmann)
    01/08/2025 11:15
    (Distributionally) robust optimization
    Contributed talk

    Distributionaly robust optimization with Wassersein-distance uncertainty sets proves to be an outstanding tool to handle data heterogeneity and distribution shifts; see (Kuhn et al.)[2].
    Recently, (Azizian et al.)[1] studied regularizations of WDRO problems. From a risk minimization problem $\min_{\theta\in\Theta}\mathbb{E}_\xi[\ell_\theta(\xi)]$ (ERM), it provides a dual formula (WDRO)...

    Aller à la page de la contribution
  215. Ivan Eryganov (Institute of Mathematics, Faculty of Mechanical Engineering, Brno University of Technology)
    01/08/2025 11:15
    Applications in energy, finance or logistics
    Contributed talk

    Waste-to-energy (WtE) plants offer a way of treating waste while converting it to energy. This provides a more sustainable way of treating waste than common landfills. A vital part of a well-functioning waste management environment is the right price setting of gate fees, i.e., treatment price per amount of waste, for the WtE plants. The price setting can be described as a non-cooperative game...

    Aller à la page de la contribution
  216. Michal Kaut (SINTEF)
    01/08/2025 11:30
    Multihorizon Stochastic Programming: Models, Algorithms, and Applications
    Invited talk

    We show how to implement long-term storage in the multi-horizon modelling paradigm, expanding the range of problems this approach is applicable to. The presented implementation is based on the HyOpt optimization model, but the ideas are generic.
    We illustrate the effects of several formulations on a simple case of electrification of an offshore installation using wind turbines and a...

    Aller à la page de la contribution
  217. Dr Mathieu Blondel (Google)
    01/08/2025 11:30
    Contributed talk
  218. Simon Thomä (RWTH Aachen)
    01/08/2025 11:45
    (Distributionally) robust optimization
    Contributed talk

    Multi-stage decision-making under uncertainty, where decisions are taken under sequentially revealing uncertain problem parameters, is often essential to faithfully model managerial problems. Given the significant computational challenges involved, these problems are typically solved approximately. This short note introduces an algorithmic framework that revisits a popular approximation scheme...

    Aller à la page de la contribution
  219. Zhuo Zhang
    01/08/2025 11:45
    Stochastic integer programming
    Invited talk

    Staged alert systems have been successfully implemented to minimize socioeconomic losses while avoiding overwhelming healthcare systems. Optimizing such systems can be formulated as a challenging two-stage stochastic mixed-integer programming problem with a discontinuous recourse function, where decision variables reside in a discrete space. Traditional simulation-based optimization techniques...

    Aller à la page de la contribution
  220. Linda Ben Ismail
    01/08/2025 11:45
    Contextual stochastic programming
    Contributed talk

    This work addresses the design and planning of a two-echelon omnichannel distribution network under endogenous uncertainty. The network consists of suppliers, retail stores, and distribution centers (DC) that serve a dual role i.e., replenishing retail stores and fulfilling direct-to-customer deliveries. We focus on strategic decisions such as the opening and configuration of distribution...

    Aller à la page de la contribution
  221. Alessandra Rende (Department of Mechanical, Energy and Management Engineering, University of Calabria, Italy)
    01/08/2025 11:45
    Applications in energy, finance or logistics
    Contributed talk

    In modern energy systems, electricity and natural gas markets are increasingly interdependent due to the prominent role of gas-fired power generation, which provides essential flexibility to balance the variability of renewable energy sources.
    In such a context, this work presents a tri-level optimization model to address the optimal bidding problem faced by a price-maker electricity...

    Aller à la page de la contribution
  222. Lukáš Račko (Charles university)
    01/08/2025 11:45
    Stochastic Programming
    Contributed talk

    In this paper, we deal with the stability of stochastic programming problems that are specified by distortion risk measures. The distortion risk measure is a specific type of risk functional that is defined as the Choquet integral of a random variable with respect to distorted probability measure. The distortion of the probability measure is governed by a distortion function that encodes the...

    Aller à la page de la contribution
  223. Ruben van Beesten (Erasmus University Rotterdam)
    01/08/2025 11:45
    Sequential decision making under uncertainty
    Invited talk

    We develop a new stopping rule for the stochastic dual dynamic programming (SDDP) algorithm based on sequential testing. Traditional stopping rules are based on statistical tests that assess whether the optimality gap in a given iteration is smaller than a pre-defined precision. However, repeated use of such single-iteration tests invalidates the overall validity of the stopping rule, because...

    Aller à la page de la contribution
  224. Giovanni Micheli (University of Bergamo)
    01/08/2025 11:55
    Multihorizon Stochastic Programming: Models, Algorithms, and Applications
    Invited talk

    In this talk we address the challenge of designing optimal domestic renewable energy systems under multiple sources of uncertainty appearing at different time scales. Long-term uncertainties, such as investment and maintenance costs of different technologies, are combined with short-term uncertainties, including solar radiation, electricity prices, and uncontrolled load variations.
    We...

    Aller à la page de la contribution
  225. Axel Parmentier (CERMICS, École Nationale des Ponts et Chaussées)
    01/08/2025 11:55
    Structured learning and stochastic combinatorial optimization: methodological perspectives and applications
    Invited talk

    As industry seeks to make its processes more resilient, data driven optimization is gaining momentum in Operation Research. Resilience involves managing uncertainty when optimizing supply chains, while efficiency requires scalable combinatorial optimization algorithms, as most gains from operations research algorithms come from decreasing marginal costs. This underscores the need for scalable...

    Aller à la page de la contribution
  226. Diego Fonseca Valero (Universidad EAFIT)
    01/08/2025 12:15
    (Distributionally) robust optimization
    Contributed talk

    We propose a novel Wasserstein distributionally robust optimization (DRO) framework with regularization control that naturally leads to a family of regularized problems with user‐controllable penalization mechanisms. Our approach bridges the gap between conventional DRO formulations and practical decision-making by explicitly incorporating adverse scenario information into the optimization...

    Aller à la page de la contribution
  227. Julio Deride (Universidad Adolfo Ibáñez)
    01/08/2025 12:15
    Stochastic Programming
    Contributed talk

    In stochastic programming, solutions are highly sensitive to approximations—whether from sampling, scenario reduction, or parametric perturbations—especially in nonconvex settings. This work investigates how substitute problems, constructed via Rockafellian functions, can provide robustness against such stochastic approximations. Unlike classical stability analysis centered on local...

    Aller à la page de la contribution
  228. Tomas Valencia Zuluaga (Lawrence Livermore National Laboratory)
    01/08/2025 12:15
    Applications in energy, finance or logistics

    Ensuring the reliability and resilience of the modern power grid requires models that handle the inherent uncertainty of generation availability and electricity consumption. These models must have sufficiently high spatial and temporal resolution to adequately capture weather variability and provide actionable siting and sizing decisions. A stochastic nodal capacity expansion planning (CEP)...

    Aller à la page de la contribution
  229. Francis Durand (Université Sorbonne Paris Nord)
    01/08/2025 12:15
    Sequential decision making under uncertainty
    Invited talk

    This work addresses the challenges of applying Stochastic Dual Dynamic Programming (SDDP) to infinite-horizon hydroelectric water management problems with continuous state and control spaces. While SDDP has proven effective in finite-horizon settings, its extension to the infinite-horizon case with a discount factor close to one introduces numerical difficulties when the discount rate is close...

    Aller à la page de la contribution
  230. Dr Hongyu Zhang (University of Southampton)
    01/08/2025 12:20
    Multihorizon Stochastic Programming: Models, Algorithms, and Applications
    Invited talk

    This paper proposes a novel algorithm to efficiently solve large-scale multi-stage stochastic programs with block separable recourse. We use an extended Benders decomposition with adaptive oracles to decompose the problem into a master problem and a collection of subproblems. The adaptive oracles enable cut sharing among the subproblems. We apply the proposed method for solving power system...

    Aller à la page de la contribution
  231. José Ignacio Delgado Anguita (Universidad Técnica Federico Santa María)
    01/08/2025 14:00
    Applications in energy, finance or logistics

    We study a robust optimization framework for the optimal operation and sizing of a hybrid virtual power plant (VPP) composed of a thermal generator unit (TGU), a large-scale photovoltaic (PV) plant, and an energy storage system (ESS). The VPP participates as a price-taker in both the day-ahead (DA) and real-time (RT) electricity markets, offering energy and ancillary services (i.e., frequency...

    Aller à la page de la contribution
  232. Sajad Khodadadian (Virginia Tech)
    01/08/2025 14:00
    Machine learning
    Invited talk
    ML

    Polyak averaging is a well-known technique for achieving asymptotically optimal convergence in Stochastic Approximation. In this work, we establish the first high-probability bound for general Stochastic Approximation with Polyak Averaging. We take a black-box approach, assuming access to an anytime high-probability bound for a given Stochastic Approximation, and derive tight finite-time...

    Aller à la page de la contribution
  233. Berend Markhorst (CWI)
    01/08/2025 14:00
    Stochastic integer programming
    Contributed talk

    Solving large-scale stochastic optimization problems is challenging, especially when uncertainties are represented by a large number of scenarios. To tackle this, we introduce TULIP ("Two-step warm start method Used for solving Large-scale stochastic mixed-Integer Problems"), an efficient approach for solving two-stage stochastic (mixed) integer programs with an exponential number of...

    Aller à la page de la contribution
  234. Vittorio Moriggia (Université de Bergame)
    01/08/2025 14:00
    Applications in energy, finance or logistics
    Contributed talk

    A Dynamic Stochastic Programming model applied to long-term Asset and Liability Management portfolio selection faces the challenge to satisfy an investor’s personal goals. Since not all the targets have the same priority, we ask the model to take the investor’s expectations into account.
    These kinds of problems are of particular interest to the insurance industry, where they are commonly...

    Aller à la page de la contribution
  235. Heiko Hoppe (Technical University of Munich)
    01/08/2025 14:00
    Structured learning and stochastic combinatorial optimization: methodological perspectives and applications
    Invited talk
    ML

    When facing contextual multi-stage optimization problems, training combinatorial optimization-enriched machine learning pipelines (ML-CO-pipelines) to date either requires imitating expert solutions or utilizing unstructured learning algorithms. While the former restricts the use of ML-CO-pipelines to problems with traceable offline solutions and relatively homogenous state spaces, the latter...

    Aller à la page de la contribution
  236. Ricardo Fukasawa (University of Waterloo)
    01/08/2025 14:30
    Stochastic integer programming
    Contributed talk

    We introduce a novel technique that generates Benders cuts from a corner relaxation of the target higher-dimensional polyhedron. Moreover, we develop a computationally efficient method to separate facet-defining inequalities for the epigraph
    associated with this corner relaxation. By leveraging a known connection between arc-flow and path-flow formulations, we show that our method can recover...

    Aller à la page de la contribution
  237. Rose Sossou Edou (École nationale des ponts et chaussées)
    01/08/2025 14:30
    Applications in energy, finance or logistics
    Invited talk

    Consider an electric battery whose owner has committed with a transmission system operator (e.g., RTE in France) to making at its disposal a "reserve'" of electricity along a given day: every hour, a random quantity will be discharged from or charged to the battery; the commitment is on the range of this random variable. To respond at best to this commitment, the owner of the battery can buy...

    Aller à la page de la contribution
  238. Jörgen Blomvall (Linköping university)
    01/08/2025 14:30
    Applications in energy, finance or logistics
    Contributed talk

    Interest rates on financial markets are noisy. This is reduced by a Kalman filter, which gives better measurement of interest rate cuts and hikes. With an optimization model interest rate curves are measured with increased accuracy from Overnight Index Swaps. Principal Component Analysis identifies the significant risk factors in interest rate markets. With these a Stochastic Programming model...

    Aller à la page de la contribution
  239. Solène Delannoy-Pavy (CERMICS / RTE)
    01/08/2025 14:30
    Structured learning and stochastic combinatorial optimization: methodological perspectives and applications
    Invited talk
    ML

    Multistage stochastic optimization (MSO) is pivotal for sequential decision-making under uncertainty, with prominent approaches in stochastic optimal control and reinforcement learning. While methods like Stochastic Dual Dynamic Programming excel with moderate-dimensional states and large continuous control spaces, and reinforcement learning handles large state spaces with smaller control...

    Aller à la page de la contribution
  240. Ilyas Fatkhullin (ETH Zürich)
    01/08/2025 14:30
    Machine learning
    Invited talk
    ML

    Federated learning faces severe communication bottlenecks due to the high dimensionality of model updates. Communication compression with contractive compressors (e.g., Top-K) is often preferable in practice but can degrade performance without proper handling. Error feedback (EF) mitigates such issues but has been largely restricted to smooth, unconstrained problems, limiting its real-world...

    Aller à la page de la contribution
  241. Laurens Elderhorst (University of Groningen)
    01/08/2025 15:00
    Stochastic integer programming
    Contributed talk

    We present a general price-and-cut procedure for two-stage stochastic integer programs with complete recourse, and first-stage binary variables. We propose a novel set of cutting planes that can close the optimality gap when added to the Dantzig-Wolfe decomposition master problem. It is shown to provide a finite exact algorithm for a number of stochastic integer programs, even in the
    presence...

    Aller à la page de la contribution
  242. Léo Baty (Ecole des Ponts)
    01/08/2025 15:00
    Structured learning and stochastic combinatorial optimization: methodological perspectives and applications
    Invited talk
    ML

    Network schedule optimization is one of the main applications of OR to the air transport industry. This process involves designing aircraft schedules that are both operationally feasible and cost-optimal. Tail assignment is the process of assigning specific aircraft (tails) to planned flights, typically occurring a few days to weeks before operations. Airlines routinely solve this problem...

    Aller à la page de la contribution
  243. Lucas Guerreiro (PSR)
    01/08/2025 15:00
    Applications in energy, finance or logistics
    Contributed talk

    This work presents a methodology for incorporating risk measures into the energy system expansion planning process. The approach involves a decomposed investment and operation model, where the objective function is modified to progressively place greater emphasis on minimizing the risk metric. By doing so, it is possible to calculate risk levels for different optimal expansion plans. The...

    Aller à la page de la contribution
  244. Honglin Wen (Shanghai Jiao Tong University)
    01/08/2025 15:00
    Applications in energy, finance or logistics
    Contributed talk

    Energy scheduling is typically conducted as a two-stage procedure that comprises day-ahead and real-time stages. Particularly, day-ahead decisions are made under uncertainty, which then affects real-time decisions, as operators must constantly maintain power balance. For that, it is commonplace in the energy industry to predict renewable energy production and power load, which are then used to...

    Aller à la page de la contribution
  245. Caleb Ju (Georgia Tech)
    01/08/2025 15:00
    Machine learning
    Invited talk
    ML

    We propose a novel termination criterion, termed the advantage gap function, for finite state and action Markov decision processes (MDP) and reinforcement learning (RL). By incorporating this advantage gap function into the design of step size rules and deriving a new linear rate of convergence that is independent of the stationary state distribution of the optimal policy, we demonstrate that...

    Aller à la page de la contribution
  246. Marius Roland (Inria Lille)
    Stochastic Programming
    Contributed talk

    Two-stage stochastic programs with finite support are a fundamental tool for decision-making under uncertainty. However, their computational tractability is often limited by the number of scenarios considered. To address this issue, scenario clustering methods have been proposed to reduce the problem size while preserving the essential characteristics of the uncertain parameters that drive the...

    Aller à la page de la contribution
  247. Samuel Affar (University of Tennessee, Knoxville)
    Sequential decision making under uncertainty
    Contributed talk

    Tri-level defender-attacker game models are a well-studied method for
    determining how best to protect a system (e.g., a transportation network) from attacks.
    Existing models assume that defender and attacker actions have a perfect effect, i.e.,
    system components hardened by a defender cannot be destroyed by the attacker, and
    attacked components always fail. Because of these assumptions,...

    Aller à la page de la contribution
  248. Guzin Bayraksan (Ohio State University)
    Applications in energy, finance or logistics
    Contributed talk

    Climate variability is increasingly affecting energy generation, particularly renewable sources that depend directly on weather patterns. Variations in temperature, precipitation, and wind patterns are altering the availability and reliability of hydropower, wind, and solar energy, posing challenges for power system stability and long-term planning. As climate patterns shift, our perception of...

    Aller à la page de la contribution
  249. Lénaïc Chizat (EPFL)
    Machine learning
    Invited talk
    ML

    Sinkhorn’s algorithm is the go-to method for solving large-scale optimal transport problems. While its theoretical foundations are rich and still expanding, a blind spot has persisted regarding the widely used heuristic of ϵ-scaling (temperature annealing), which lacks convergence guarantees. In this talk I will present such guarantees as well as theoretical insights into the design of...

    Aller à la page de la contribution
  250. Vít Procházka (Charles University)
    Applications in energy, finance or logistics
    Contributed talk

    We discuss a new electric car ferry operating along the west coast of Norway. The ferry visits a total of 17 different ports, of which only three are mandatory stops. It can carry up to 27 standard cars across three lanes. Depending on the loading arrangement, cars may need to reverse onto or off the ferry. In some cases, vehicles may also need to disembark at an intermediate port and reboard...

    Aller à la page de la contribution
  251. Dr Abolhassan Mohammadi Fathabad (Postdoc at the School of Industrial Engineering, University of Tehran)
    (Distributionally) robust optimization
    Contributed talk

    It is often the case where historical data used to represent uncertainty need some expert-based opinion before being suitable for use in a planning problem. For instance, in a renewable energy and energy storage planning (REESP) problem, the solar generation data may be gathered using older generation of solar panel technology while the nowel solar panels intended for planning have improved...

    Aller à la page de la contribution
  252. Rohit Kannan (Virginia Tech)
    Contextual Stochastic Programming
    Invited talk

    We propose a data-driven framework for multistage stochastic optimization using only a single historical trajectory of the underlying process. Given recent observations of the stochastic process, the goal is to find a policy that minimizes expected cost over the next T time periods. Our approach fits a time-series model to the data and uses its residuals to construct a discrete approximation...

    Aller à la page de la contribution
  253. Francis Bach
    Plenary talk

    Denoising diffusion models have enabled remarkable advances in generative modeling across various domains. These methods rely on a two-step process: first, sampling a noisy version of the data—an easier computational task—and then denoising it, either in a single step or through a sequential procedure. Both stages hinge on the same key component: the score function, which is closely tied to...

    Aller à la page de la contribution
  254. Jiaxin Shi (Google DeepMind)
    Machine learning
    Invited talk
    ML

    Modern generative AI has developed along two distinct paths: autoregressive models for discrete data (such as text) and diffusion models for continuous data (like images). Bridging this divide by adapting diffusion models to handle discrete data represents a compelling avenue for unifying these disparate approaches. However, existing work in this area has been hindered by unnecessarily complex...

    Aller à la page de la contribution
  255. Jia Liu (Xi'an Jiaotong University)
    Stochastic Programming
    Contributed talk

    We investigate the elicitation method for the Von Neumann–Morgenstern-type decision-maker (DM) from pairwise comparison data in the presence of response errors. We apply the maximum likelihood estimation (MLE) method to elicit the nominal utility, together with the variance of the response error, assuming a Gumbel distribution. Given the finite support of the pairwise comparison lotteries and...

    Aller à la page de la contribution
  256. Zhengqi Lin
    Machine learning
    Invited talk
    ML

    We propose an efficient federated dual decomposition algorithm for calculating the
    Wasserstein barycenter of several distributions. The algorithm does not have
    access to local data and uses only highly aggregate information. Owing to the
    absence of any matrix-vector operations, the algorithm exhibits very low complexity of
    each iteration and significant scalability. We illustrate its...

    Aller à la page de la contribution
  257. Jincheng Mei (Google DeepMind)
    Machine learning
    Invited talk
    ML

    We provide a new understanding of the stochastic gradient bandit algorithm by showing that it converges to a globally optimal policy almost surely using any constant learning rate. This result demonstrates that the stochastic gradient algorithm continues to balance exploration and exploitation appropriately even in scenarios where standard smoothness and noise control assumptions break...

    Aller à la page de la contribution
  258. Dr Quentin Berthet (Google)

    We present a new algorithm to optimize distributions defined implicitly by parameterized stochastic diffusions. Doing so allows us to modify the outcome distribution of sampling processes by optimizing over their parameters. We introduce a general framework for first-order optimization of these processes, that performs jointly, in a single loop, optimization and sampling steps. This approach...

    Aller à la page de la contribution
  259. Jakub Marecek (Czech Technical University in Prague)
    Stochastic integer programming
    Contributed talk

    Optimization under uncertainty often relies on sampling the uncertainty to evaluate some expectations or to work with probability constraints. As the dimension of the random variable grows, the applicability of this approach diminishes, as achieving non-trivial precision through sampling may require too many samples to be practical.
    We propose an alternative approach, where the probability...

    Aller à la page de la contribution
  260. Mathieu Dahan (Georgia Institute of Technology)
    Applications in energy, finance or logistics
    Contributed talk

    We study the design of large-scale relay logistics hub networks that are resilient to demand variability. We formulate a two-stage stochastic optimization model that integrates first-stage strategic decisions on hub location and capacity with second-stage tactical decisions on consolidation-based routing. To solve this problem exactly, we develop a three-stage branch-and-cut algorithm with...

    Aller à la page de la contribution
  261. Soroosh Shafiee
    Contextual stochastic programming
    Invited talk

    We study learning in an adversarial setting, where an epsilon fraction of samples from a distribution P are globally corrupted (arbitrarily modified), and the remaining perturbations have an average magnitude bounded by rho (local corruptions). With access to n such corrupted samples, we aim to develop a computationally efficient approach that achieves the optimal minimax excess risk. Our...

    Aller à la page de la contribution
  262. Cagil Kocyigit (University of Luxembourg)
    Machine learning
    Invited talk
    ML

    We study the problem of learning a treatment assignment policy based on observable covariates, where there are potential shifts in the distribution of covariates from historical data (training) to deployment (test). We formulate a distributional robust policy optimization problem with the objective of maximizing the worst-case (out-of-sample) expected outcomes, considering all possible...

    Aller à la page de la contribution
  263. Zhi Chen (The Chinese University of Hong Kong)
    Computationally Efficient Approaches for Distributionally Robust Optimization
    Invited talk

    We study a multidimensional mechanism design problem where a seller offers multiple products to a single buyer. The seller possesses only marginal distributional information about the buyer's random valuation of each product. The buyer is not a perfect optimizer and is satisficing whenever his incentive is epsilon away from the optimal---a notion called approximate incentive compatibility...

    Aller à la page de la contribution
  264. Egor Shulgin (King Abdullah University of Science and Technology (KAUST))
    Machine learning
    Contributed talk
    ML

    Federated learning enables training machine learning models while preserving the privacy of participants. Surprisingly, there is no differentially private distributed method for smooth, non-convex optimization problems. The reason is that standard privacy techniques require bounding the participants' contributions, usually enforced via clipping of the updates. Existing literature typically...

    Aller à la page de la contribution