Changeset 11770 for branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/GrammarPolicies/RandomPolicy.cs

Timestamp:

01/15/15 18:59:07 (9 years ago)

Author:

gkronber

Message:

#2283: worked on generic sequential search alg with bandit policy as parameter

File:

• branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/GrammarPolicies/RandomPolicy.cs (modified) (1 diff)

branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/GrammarPolicies/RandomPolicy.cs

@@ -8 +8 @@
 
 namespace HeuristicLab.Algorithms.Bandits.GrammarPolicies {
-  public class RandomPolicy : IGrammarPolicy {
-    public ReadonlySequence SelectAction(Random random, ReadonlySequence state, IEnumerable<ReadonlySequence> actions) {
-      return actions.SelectRandom(random);
+  public class RandomPolicy : GrammarPolicy {
+    public RandomPolicy(IProblem problem, bool useCanonicalRepresentation)
+      : base(problem, useCanonicalRepresentation) {
     }
 
-    public void UpdateReward(ReadonlySequence state, ReadonlySequence action, double reward, ReadonlySequence newState) {
-      // ignore
-    }
-
-    public bool Done(ReadonlySequence state) {
-      return false;
+    public override bool TrySelect(Random random, ReadonlySequence curState, IEnumerable<ReadonlySequence> afterStates, out ReadonlySequence selectedState) {
+      // never fail => allows re-visits of terminal states
+      selectedState = afterStates.SelectRandom(random);
+      return true;
     }
   }