TITLE: Selected Topics in Planning under Uncertainty
SPEAKER: Boris Defourny (Systems and Modeling Research Unit, University of Liège)

ABSTRACT:

The first part of the talk concerns sequential decision making under
uncertainty and methods that use dynamic programming to find optimal
policies. We will discuss how Bellman's principle limits the ways
risk-awareness can be incorporated in sequential decision making, but also
brings time-consistency.

The second part of the talk concerns sequential decision making under
uncertainty and methods that use stochastic programming to find optimal
policies. The example of the newsvendor problem will serve to introduce
the stochastic programming approach, and the issues raised by
finite-dimensional approximations to stochastic programs.
Then, I will present some recent work applying machine learning techniques
and scenario-tree techniques from multistage stochastic programming to
sequential decision making.