SEMINARS
Stochastic algorithms for sampling high-dimensional Boltzemann measures

2019-07-29　16:00 — 17:00

Middle Lecture Room(703)

Liming Wu

Centre national de la recherche scientifique

The integral of a function w.r.t. a high dimensional Boltzeman measure $\mu=(1/C)e^{-V} dx$ is impossible in deterministic approaches. In this survey talk I will present several old and current stochastic algorithms : Gibbs sampling, Langevin's equation or Metropolis-Hasting. However many mathematical problems remain open, which are new important chalenge for today's and future probabilists.