next up previous contents
Next: Markov Decision Problems Up: Learning Evaluation Functions Previous: Proposal

Value Function Approximation for Prediction and Control

 

The optimal value function is an evaluation function which encapsulates complete knowledge of the best expected search outcome attainable from each state:

  equation72

Such an evaluation function is ideal in that a greedy local search with respect to tex2html_wrap_inline1400 will always make the globally optimal move. This section reviews the literature on computing tex2html_wrap_inline1454 , and motivates and describes our new approximation algorithms for this problem.





Justin A. Boyan
Sat Jun 22 20:49:48 EDT 1996