NeuralNetwork.h

Go to the documentation of this file.

/*
 * NeuralNetwork.h
 *
 * A simple feedforward neural network with one hidden layer.
 *
 * This file is part of Qualia https://github.com/sofian/qualia
 *
 * (c) 2011 Sofian Audry -- info(@)sofianaudry(.)com
 * Inspired by code by Karsten Kutza
 * http://www.ip-atlas.com/pub/nap/nn-src/bpn.txt
 *
 * 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 3 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
#ifndef NEURAL_NETWORK_INC
#define NEURAL_NETWORK_INC

#include <stdlib.h>
#include <math.h>
#include <string.h>

#include <qualia/learning/GradientFunction.h>
#include <qualia/core/common.h>
#include <qualia/util/random.h>

// TODO: IMPORTANT l'output layer ne devrait pas etre sigmoide mais lineaire...

class NeuralNetwork : public GradientFunction {

public:
  // Configurable parameters /////

  float learningRate;

  float decreaseConstant;

  float weightDecay;

  // Internal use ////////////////

  float _learningRateDiv;

  struct Layer {
    unsigned int n;          // number of units in this layer
    real *output;   // output of ith unit
    real *error;    // error term of ith unit
    real *weight;   // connection weights to ith unit
    real *dWeight;  // weight derivatives
    bool linear;    // whether the output of the layer is linear (otherwise it's sigmoid)
  };

  unsigned int _nParams;

  Layer inputLayer, hiddenLayer, outputLayer;

public:
  // Interface ///////////////////

  // Constructor/destructor.
  NeuralNetwork(unsigned int nInputs,
                unsigned int nHiddens,
                unsigned int nOutputs,
                float learningRate = 0.01,
                float decreaseConstant = 0,
                float weightDecay = 0,
                bool linearOutput = false);
  virtual ~NeuralNetwork();

  // Public methods.

  virtual void init();

  virtual unsigned int nInputs() const { return inputLayer.n; }

  virtual unsigned int nHidden() const { return hiddenLayer.n; }

  virtual unsigned int nOutputs() const { return outputLayer.n; }

  virtual unsigned int nParams() const { return _nParams; }

  virtual float getCurrentLearningRate() const;

  virtual void setInput(int i, real x);
  virtual void setInputs(const real *inputs);

  virtual real getOutput(int i) const;
  virtual void getOutputs(real *outputs) const;

  virtual void backpropagate(real *outputError);

  virtual void propagate();

  virtual void update();

  virtual void save(XFile* file);
  virtual void load(XFile* file);

#ifdef DEBUG

  void printLayer(Layer* layer, Layer* lower);

  void debug();
#endif

  // Internal ("private") methods.
  void _allocateLayer(Layer& layer, unsigned int nInputs, unsigned int nOutputs, unsigned int& k, bool isLinear=false);
  void _deallocateLayer(Layer& layer);

  void _propagateLayer(Layer& lower, Layer& upper);
  void _backpropagateLayer(Layer& upper, Layer& lower);
  void _deallocate();
};

#endif