COLT
2000
Springer
14 years 3 months ago
2000
Springer
We give the first rigorous upper bounds on the error of temporal difference (td) algorithms for policy evaluation as a function of the amount of experience. These upper bounds pr...