Bayesian Reinforcement Learning for Problems with State Uncertainty - VVSOR - VVSOR

Netherlands Society for Statistics and Operations Research | Dutch
01 March 2019

Bayesian Reinforcement Learning for Problems with State Uncertainty

We cordially invite you to the next meeting of the Thematic Statistics Seminar with current focus on Machine Learning from a statistical perspective on Friday, March 1 in Leiden:

Speaker: Frans Oliehoek (TU Delft)
Title: Bayesian Reinforcement Learning for Problems with State Uncertainty
Time: 15:00-16:00, March 1, 2019
Location: Room 405, Snellius Building, Leiden University, Niels Bohrweg 1, Leiden

Abstract: Sequential decision making under uncertainty is a challenging problem, especially when the decision maker, or agent, has uncertainty about what the true ‘state’ of the environment is. The partially observable Markov decision process (POMDP) framework models such problem that are ‘partially observable’: there are important pieces of information that are fundamentally hidden from the agent. Planning in a POMDP is difficult, but the problem gets even more complex when no accurate model of the environment is available. In such cases, the agent will need to update its belief over the environment, i.e., learn, during execution.

In this talk, I will introduce the POMDP framework, as well as a more recent extension to the learning setting called Bayes-Adaptive POMDP (BA-POMDP). I will explain how the learning problem can be tackled using a Monte Carlo tree search method called ‘POMCP’, how this can be made more efficient via a number of novel techniques, and how we can further increase its effectiveness by exploiting structure in the environment. Time permitting, I will also discuss extensions of this methodology that explicitly deal with coordination with other agents, and anticipation of other actors (such as humans) in the environment.

For the list of upcoming talks and further information about the seminar please visit the seminar webpage: