BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Provably Efficient Supervised Learning Algorithms for Non-i.i.d. a
 nd Structured Data
DTSTART:20251007T091500Z
DTEND:20251007T101500Z
DTSTAMP:20260816T194700Z
UID:indico-event-14450@indico.math.cnrs.fr
DESCRIPTION:Speakers: Luc Brogat-Motte (Istituto Italiano di Tecnologia (I
 IT))\n\nThis talk will present recent contributions and ongoing directions
  in supervised learning for non-i.i.d. and structured data. It will presen
 t algorithms that are both efficient in practice and come with formal perf
 ormance guarantees. These approaches are available as open-source software
  and have been validated on both synthetic benchmarks and real-world datas
 ets. The first part will focus on structured prediction methods designed t
 o overcome the curse of the output dimension\, providing finite-sample lea
 rning bounds and computational complexity guarantees\, and demonstrating s
 ignificant empirical improvements over state-of-the-art approaches in appl
 ications such as metabolite identification. The second part will address l
 earning from time-dependent data\, combining statistical learning theory w
 ith stochastic calculus to develop estimation methods for continuous-time\
 , nonlinear stochastic dynamical systems. Particular attention will be giv
 en to stochastic differential equation estimation and to safe learning for
  controlled systems under uncertainty\, with applications in areas such as
  robotics and biology.\n\nhttps://indico.math.cnrs.fr/event/14450/
LOCATION:Salle K. Johnson (1R3\, 1er étage)
URL:https://indico.math.cnrs.fr/event/14450/
END:VEVENT
END:VCALENDAR