Changeset 11732 for branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GrammaticalOptimization/AlternativesContextSampler.cs

Timestamp:

01/07/15 09:21:46 (9 years ago)

Author:

gkronber

Message:

#2283: refactoring and bug fixes

File:

• branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GrammaticalOptimization/AlternativesContextSampler.cs (modified) (5 diffs)

branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GrammaticalOptimization/AlternativesContextSampler.cs

@@ -17 +17 @@
     private readonly Random random;
     private readonly int contextLen;
-    private readonly Func<Random, int, IPolicy> policyFactory;
+    private readonly IPolicy policy;
 
-    public AlternativesContextSampler(IProblem problem, Random random, int maxLen, int contextLen, Func<Random, int, IPolicy> policyFactory) {
+    public AlternativesContextSampler(IProblem problem, Random random, int maxLen, int contextLen, IPolicy policy) {
       this.maxLen = maxLen;
       this.problem = problem;
       this.random = random;
       this.contextLen = contextLen;
-      this.policyFactory = policyFactory;
+      this.policy = policy;
     }
 
@@ -32 +32 @@
       for (int i = 0; i < maxIterations; i++) {
         var sentence = SampleSentence(problem.Grammar).ToString();
-        var quality = problem.Evaluate(sentence) / problem.GetBestKnownQuality(maxLen);
+        var quality = problem.Evaluate(sentence) / problem.BestKnownQuality(maxLen);
         DistributeReward(quality);
 
@@ -45 +45 @@
 
 
-    private Dictionary<string, IPolicy> ntPolicy;
+    private Dictionary<string, IPolicyActionInfo[]> contextActionInfos;
     private List<Tuple<string, int>> updateChain;
 
     private void InitPolicies(IGrammar grammar) {
-      this.ntPolicy = new Dictionary<string, IPolicy>();
+      this.contextActionInfos = new Dictionary<string, IPolicyActionInfo[]>();
       this.updateChain = new List<Tuple<string, int>>();
     }
@@ -83 +83 @@
           var lft = phrase.Subsequence(startIdx, endIdx - startIdx + 1).ToString();
           lft = problem.Hash(lft);
-          if (!ntPolicy.ContainsKey(lft)) {
-            ntPolicy.Add(lft, policyFactory(random, g.GetAlternatives(nt).Count()));
+          if (!contextActionInfos.ContainsKey(lft)) {
+            contextActionInfos.Add(lft, g.GetAlternatives(nt).Select(_ => policy.CreateActionInfo()).ToArray());
           }
-          var selectedAltIdx = ntPolicy[lft].SelectAction();
+          var selectedAltIdx = policy.SelectAction(random, contextActionInfos[lft]);
           selectedAlt = alts.ElementAt(selectedAltIdx);
           updateChain.Add(Tuple.Create(lft, selectedAltIdx));
@@ -103 +103 @@
         var lft = e.Item1;
         var action = e.Item2;
-        ntPolicy[lft].UpdateReward(action, reward);
+        contextActionInfos[lft][action].UpdateReward(reward);
       }
     }