// evaluation_alg.h :
// generic algorithm definition for algoritms for value function
// (per state) experimental evaluation.

template  <class State,class Action>
class Evaluation_Alg : public Learning_Alg<State,Action> {

protected:

  State         CurrState;  
  int           RunCount;   // number of runs performd so far
  int           StepCount;
  StateStats_M  ValuesFunc; //  statistics accumulated so fary
  double        Lambda;

  void check_in_table (State newState) {
    // sets entry to 0 if it's not exists already

    const StateStats_M& valuesData = ValuesFunc;

    StateStats_I nextStepValI = valuesData.find (newState);
  
    if (nextStepValI == valuesData.end()) {
      ValuesFunc[newState] = 0;
    }
  }

public:
  
  Evaluation_Alg(double lambdaVal = 1.0) {
    RunCount  = 0;
    StepCount = 0;
    Lambda    = lambdaVal;

  }

  // start new experiment in <startState> state
  virtual void start_run (State startState) {

    CurrState = startState;
    RunCount++;
  }
  
  // just prints values function estimation
  virtual void print_stats () {
    const StateStats_M& valuesData = ValuesFunc;

    cout << RunCount << " runs\n";

    for (StateStats_I allVals = valuesData.begin();
         allVals != valuesData.end();
         ++allVals)

      cout << (*allVals).first << " : " << (*allVals).second << "\n";

    cout << "\n";

  }


};