## News

### Lectures Notice

**Tuesday 20/12/2011, 2pm, Seminar Room. Alan Perotti and Francesco Osborne will present two lectures as final examination for the PhD course**

*Intelligent Decision Support Systems***Lecture 1**: Alan Perotti*Abstract: *"Probability propagation" by Shafer and Shenoy and "Morphing the Hugin and Shenoy/Shafer Architectures" by Park and Darwiche. ** **

**Lecture 2: **Francesco Osborne*Abstract*: the Partially Observable Markov Decision Problem (POMDP) is an extension of the MDP, which provides a mathematical framework for modeling sequential decision-making under uncertainty. In fact a POMDP models an agent decision process in a context in which system states are not fully observable and thus the agent must maintain a probability distribution over the set of possible state. An exact solution to a POMDP yields the actions which optimize the "reward" for each possible belief over the world states.

However, solving a POMDP exactly is often a virtually intractable problem, since an agent might need to take into account all the previous history of observations and actions. For this reason various techniques that approximate solutions for POMDPs have been developed.