Post contrassegnati da tagMDP

Pythonic Markov Decision Process (MDP)

I’ve been studying a subject called probabilistic methods for decisions.
In this course there are a lot of interesting topics like Bayesian Networks (BN), inference and querying in BNs, probabilistic reasoning over time etc. and, in the book (Artificial Intelligence: A Modern Approach - Russell, Norvig) there are a lot of interesting pseudocode ;) to implement.

The simplest algorithm to implement is -IMHO- the Value-Iteration algorithm and my goal is to reproduce the graph that shows the evolution of the utilities shown in Figure 17.5.

MDP example
(image source)

Leggi il seguito di questo post »

Commenti