Changeset 11755 for branches/HeuristicLab.Problems.GrammaticalOptimization/Main/Program.cs

Timestamp:

01/13/15 20:02:29 (10 years ago)

Author:

gkronber

Message:

#2283: implemented synthetic benchmark problems (modeling symb-reg) with configurable hardness

File:

• branches/HeuristicLab.Problems.GrammaticalOptimization/Main/Program.cs (modified) (3 diffs)

branches/HeuristicLab.Problems.GrammaticalOptimization/Main/Program.cs

@@ -146 +146 @@
       // TODO: wie kann ich sampler noch vergleichen bzw. was kann man messen um die qualität des samplers abzuschätzen (bis auf qualität und iterationen bis zur besten lösung) => ziel schnellere iterationen zu gutem ergebnis
       // TODO: research thompson sampling for max bandit?
-      // TODO: ausführlicher test von strategien für k-armed max bandit
+      // TODO: ausführlicher test von strategien für numCorrectPhrases-armed max bandit
       // TODO: verify TA implementation using example from the original paper      
       // TODO: separate policy from MCTS tree data structure to allow sharing of information over disconnected parts of the tree (semantic equivalence)
@@ -166 +166 @@
       var random = new Random();
 
-      var phraseLen = 1;
-      var sentenceLen = 25;
-      var numPhrases = sentenceLen / phraseLen;
-      var problem = new RoyalPhraseSequenceProblem(random, 10, numPhrases, phraseLen: 1, k: 1, correctReward: 1, incorrectReward: 0);
-
-      //var problem = new SymbolicRegressionPoly10Problem();   // good results e.g. 10 randomtries and EpsGreedyPolicy(0.2, (aInfo)=>aInfo.MaxReward)
+      //var phraseLen = 3;
+      //var numPhrases = 5;
+      //var problem = new RoyalPhraseSequenceProblem(random, 10, numPhrases, phraseLen: phraseLen, numCorrectPhrases: 1, correctReward: 1, incorrectReward: 0.0, phrasesAsSets: true);
+
+      //var phraseLen = 4;
+      //var numPhrases = 5;
+      //var problem = new FindPhrasesProblem(random, 15, numPhrases, phraseLen, numOptimalPhrases: numPhrases, numDecoyPhrases: 500, correctReward: 1.0, decoyReward: 0.2, phrasesAsSets: true);
+
+      var problem = new SymbolicRegressionPoly10Problem();   // good results e.g. 10 randomtries and EpsGreedyPolicy(0.2, (aInfo)=>aInfo.MaxReward)
       // Ant
       // good results e.g. with       var alg = new MctsSampler(problem, 17, random, 1, (rand, numActions) => new ThresholdAscentPolicy(numActions, 500, 0.01));
@@ -182 +185 @@
       //var problem = new EvenParityProblem();
       // symbreg length = 11 q = 0.824522210419616
-      var alg = new MctsSampler(problem, sentenceLen, random, 0, new BoltzmannExplorationPolicy(200));
+      //var alg = new MctsSampler(problem, 23, random, 0, new BoltzmannExplorationPolicy(100));
+      var alg = new MctsSampler(problem, 23, random, 0, new EpsGreedyPolicy(0.1));
       //var alg = new MctsQLearningSampler(problem, sentenceLen, random, 0, null);
       //var alg = new MctsQLearningSampler(problem, 30, random, 0, new EpsGreedyPolicy(0.2));