Orateur
Description
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 representation, such as a sample average approximation, of an MSLP problem. In contrast, stochastic dynamic linear programming (SDLP) is an internal sampling algorithm that operates on a dynamically evolving representation of an MSLP, using sequentially generated sample paths. In this talk, we present a comparative study of these two prominent sampling-based approaches for solving stagewise independent MSLP problems. We also present a regularized variant of SDDP that, like SDLP, employs a basic feasible policy to identify the prox-center. We evaluate the algorithms on benchmark instances of varying sizes and structures, assessing their convergence behavior and computational efficiency. A series of numerical experiments is conducted to demonstrate the trade-offs among these methods in terms of accuracy and scalability.
