Partially Observable Markov Decision Process (POMDP)

Some parts of this note are from the slides of Stanford NLU course

In POMDP, the bot does not know the explicit state it is in. But there are some observations.

screen shot 2014-10-10 at 14 21 41

screen shot 2014-10-10 at 14 22 28

Optimization

screen shot 2014-10-10 at 14 22 59

screen shot 2014-10-10 at 14 23 35