Changeset 13492 for branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/Policies/UCTPolicy.cs

Timestamp:

01/10/16 17:13:27 (9 years ago)

Author:

aballeit

Message:

#2283 UCT parameter c

File:

• branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/Policies/UCTPolicy.cs (modified) (2 diffs)

branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.Bandits/Policies/UCTPolicy.cs

@@ -12 +12 @@
     private readonly double c;
 
-    public UCTPolicy(double c = 1.0) {
+      // c = sqrt(2)
+    public UCTPolicy(double c = 1.41421356237)
+    {
       this.c = c;
     }
@@ -30 +32 @@
           q = double.PositiveInfinity;
         } else {
-            q = aInfo.SumReward / aInfo.Tries + 2.0 * c * Math.Sqrt(Math.Log(totalTries) / aInfo.Tries);
+            q = aInfo.SumReward / aInfo.Tries + c * Math.Sqrt(Math.Log(totalTries) / aInfo.Tries);
         }
         if (q > bestQ) {