public MDPUtilityFunction <TState> PolicyEvaluation(
MDPPolicy <TState, TAction> pi,
MDPUtilityFunction <TState> u, double gamma, int iterations)
{
var uDash = u.Copy();
for (var i = 0; i < iterations; i++)
{
uDash = this.ValueIterateOnceWith(gamma, pi, uDash);
}
return(uDash);
}
private MDPUtilityFunction <TState> ValueIterateOnceWith(double gamma,
MDPPolicy <TState, TAction> pi,
MDPUtilityFunction <TState> U)
{
MDPUtilityFunction <TState> uDash = U.Copy();
foreach (var s in this.nonFinalstates)
{
var highestPolicyTransition =
this.transitionModel.GetTransitionWithMaximumExpectedUtilityUsingPolicy(pi, s, U);
double utility = rewardFunction.GetRewardFor(s)
+ (gamma * highestPolicyTransition.GetSecond());
uDash.SetUtility(s, utility);
}
// TODO: debugging code
// System.out.println("ValueIterationOnce before " + U);
// System.out.println("ValueIterationOnce after " + U_dash);
return(uDash);
}
