Fixed backtrack prediction training

trainer/src/Trainer.cpp

@@ -306,19 +306,19 @@ void Trainer::doStepTrain()
     if (TI.getEpoch() >= k && choiceWithProbability(ProgramParameters::dynamicProbability))
     {
       actionName = pAction;
-      TI.lastActionWasPredicted[tm.getCurrentClassifier()->name] = true;
+      TI.lastActionWasPredicted[trainConfig.getCurrentStateName()] = true;
     }
     else
     {
       if (pActionIsZeroCost)
       {
         actionName = pAction;
-        TI.lastActionWasPredicted[tm.getCurrentClassifier()->name] = true;
+        TI.lastActionWasPredicted[trainConfig.getCurrentStateName()] = true;
       }
       else
       {
         actionName = oAction;
-        TI.lastActionWasPredicted[tm.getCurrentClassifier()->name] = false;
+        TI.lastActionWasPredicted[trainConfig.getCurrentStateName()] = false;
       }
 
     }
@@ -367,25 +367,24 @@ void Trainer::doStepTrain()
     }
 
     actionName = pAction;
-    if (choiceWithProbability(0.6) && TI.getEpoch() >= ProgramParameters::dynamicEpoch)
+    if (TI.getEpoch() < ProgramParameters::dynamicEpoch)
+      actionName = oAction;
+    else if (actionName == "EPSILON")
       actionName = oAction;
 
     char buffer[1024];
-    if (sscanf(tm.getCurrentClassifier()->name.c_str(), "Error_%s", buffer) != 1)
+    if (sscanf(trainConfig.getCurrentStateName().c_str(), "error_%s", buffer) != 1)
     {
       fprintf(stderr, "ERROR (%s) : unexpected classifier name. Aborting.\n", ERRINFO);
       exit(1);
     }
-    std::string normalClassifierName(buffer);
+    std::string normalStateName(buffer);
 
-    auto & normalHistory = trainConfig.getActionsHistory(normalClassifierName);
+    auto & normalHistory = trainConfig.getActionsHistory(normalStateName);
 
     // If a BACK just happened
     if (normalHistory.size() > 1 && trainConfig.getCurrentStateHistory().size() > 0 && split(trainConfig.getCurrentStateHistory().top(), ' ')[0] == "BACK" && TI.getEpoch() >= ProgramParameters::dynamicEpoch)
     {
-      fprintf(stderr, "Current classifier : <%s>\n", tm.getCurrentClassifier()->name.c_str());
-      fprintf(stderr, "Current state history : <%s>\n", trainConfig.getCurrentStateHistory().top().c_str());
-
       auto & lastAction = trainConfig.lastUndoneAction;
       auto & newAction = normalHistory[normalHistory.size()-1];
       auto & lastActionName = lastAction.first;
@@ -398,7 +397,7 @@ void Trainer::doStepTrain()
         fprintf(stderr, "<%s>(%d) -> <%s>(%d)\n", lastActionName.c_str(), lastCost, newActionName.c_str(), newCost);
       }
 
-      if (TI.lastActionWasPredicted[normalClassifierName])
+      if (TI.lastActionWasPredicted[normalStateName])
       {
         if (ProgramParameters::debug)
         {