ann_policy.h

Go to the documentation of this file.

// -*- Mode: c++ -*-
// copyright (c) 2004 by Christos Dimitrakakis <dimitrak@idiap.ch>
// $Id$
 
 
/***************************************************************************
 *                                                                         *
 *   This program is free software; you can redistribute it and/or modify  *
 *   it under the terms of the GNU General Public License as published by  *
 *   the Free Software Foundation; either version 2 of the License, or     *
 *   (at your option) any later version.                                   *
 *                                                                         *
 ***************************************************************************/
 
#ifndef ANN_POLICY_H
#define ANN_POLICY_H
 
#include <learning/policy.h>
class ANN_Policy : public DiscretePolicy
{
protected:
    ANN* J; 
    ANN** Ja; 
    real* ps; 
    real* JQs; 
    real J_ps_pa; 
    real* delta_vector; 
    bool eligibility; 
    bool separate_actions; 
public:
    ANN_Policy (int n_states, int n_actions, int n_hidden = 0, real alpha=0.1, real gamma=0.8, real lambda=0.8, bool eligibility = false, bool softmax = false, real randomness=0.1, real init_eval=0.0, bool separate_actions = false);
    virtual ~ANN_Policy();
    virtual int SelectAction(int s, real r, int forced_a=-1);
    virtual int SelectAction(real* s, real r, int forced_a=-1);
    virtual void Reset();
    virtual real getLastActionValue () {return J_ps_pa;} 
    virtual real* getActionProbabilities () {
        real sum = 0.0;
        int i;
        for (i=0; i<n_actions; i++) {
            sum += eval[i];
        }
        for (i=0; i<n_actions; i++) {
            eval[i] = eval[i]/sum;
        }
        return eval;
    }

    virtual bool useConfidenceEstimates(bool confidence, real zeta, bool confidence_eligibility);
    virtual bool useConfidenceEstimates(bool confidence, real zeta=0.01);
};
 
 
#endif