Reinforcement Learning Toolbox 2.0 last updated:

General Documentation Manual Tutorial Class Reference Master Thesis Examples Related Papers Downloads Links News mailto:webmaster Main Page     Class Hierarchy   Compound List   File List   Compound Members   File Members csupervisedlearner.h Go to the documentation of this file. // Copyright (C) 2003 // Gerhard Neumann (gneumann@gmx.net) // Stephan Neumann (sneumann@gmx.net) // // This file is part of RL Toolbox. // http://www.igi.tugraz.at/ril_toolbox // // All rights reserved. // // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions // are met: // 1. Redistributions of source code must retain the above copyright // notice, this list of conditions and the following disclaimer. // 2. Redistributions in binary form must reproduce the above copyright // notice, this list of conditions and the following disclaimer in the // documentation and/or other materials provided with the distribution. // 3. The name of the author may not be used to endorse or promote products // derived from this software without specific prior written permission. // // THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR // IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES // OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. // IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, // INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT // NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF // THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. #ifndef C_SUPERVISEDTRAINER__H #define C_SUPERVISEDTRAINER__H #include "cparameters.h" class Matrix; class ColumnVector; class CGradientFunction; class CGradientUpdateFunction; class CDataSet; class CDataSet1D; class CAction; class CTorchGradientFunction; class CFeatureList; class CFeatureFunction; class CSupervisedLearner : virtual public CParameterObject { protected: public: CSupervisedLearner() {}; virtual ~CSupervisedLearner() {}; virtual void learnFA(CDataSet *inputData, CDataSet1D *outputData) = 0; virtual void resetLearner() {}; }; class CSupervisedWeightedLearner : virtual public CParameterObject { protected: public: CSupervisedWeightedLearner() {}; virtual ~CSupervisedWeightedLearner() {}; virtual void learnWeightedFA(CDataSet *inputData, CDataSet1D *outputData, CDataSet1D *weighting) = 0; virtual void resetLearner() {}; }; class CSupervisedQFunctionLearner : virtual public CParameterObject { protected: public: CSupervisedQFunctionLearner() {}; virtual ~CSupervisedQFunctionLearner() {}; virtual void learnQFunction(CAction *action, CDataSet *inputData, CDataSet1D *outputData) = 0; virtual void resetLearner() {}; }; class CSupervisedQFunctionWeightedLearner : virtual public CParameterObject { protected: public: CSupervisedQFunctionWeightedLearner() {}; virtual ~CSupervisedQFunctionWeightedLearner() {}; virtual void learnQFunction(CAction *action, CDataSet *inputData, CDataSet1D *outputData, CDataSet1D *weightingData) = 0; virtual void resetLearner() {}; }; class CLeastSquaresLearner : virtual public CParameterObject { protected: Matrix *A; Matrix *A_pinv; ColumnVector *b; CGradientUpdateFunction *featureFunc; public: CLeastSquaresLearner(CGradientUpdateFunction *featureFunc, int numData); virtual ~CLeastSquaresLearner(); virtual double doOptimization(); static double doOptimization(Matrix *A, Matrix *A_pinv, ColumnVector *b, ColumnVector *w, double lambda); }; class CGradientCalculator : virtual public CParameterObject { protected: public: virtual ~CGradientCalculator() {}; virtual void getGradient(CFeatureList *gradient) = 0; virtual double getFunctionValue() = 0; virtual void resetGradientCalculator() {}; }; class CSupervisedGradientCalculator : public CGradientCalculator { protected: CDataSet *inputData; CDataSet1D *outputData1D; CDataSet *outputData; CGradientFunction *gradientFunction; public: CSupervisedGradientCalculator(CGradientFunction *gradientFunction, CDataSet *inputData, CDataSet *outputData); virtual ~CSupervisedGradientCalculator(); virtual void getGradient(CFeatureList *gradient); virtual double getFunctionValue(); virtual void setData(CDataSet *inputData, CDataSet1D *outputData1D); virtual void setData(CDataSet *inputData, CDataSet *outputData); }; class CSupervisedFeatureGradientCalculator : public CSupervisedGradientCalculator { protected: CFeatureFunction *featureFunction; CFeatureList *featureList; public: CSupervisedFeatureGradientCalculator(CFeatureFunction *featureFunction); virtual ~CSupervisedFeatureGradientCalculator(); CFeatureList *getFeatureList(ColumnVector *input); virtual void getGradient(CFeatureList *gradient); virtual double getFunctionValue(); }; class CGradientFunctionUpdater : virtual public CParameterObject { protected: CGradientUpdateFunction *updateFunction; public: CGradientFunctionUpdater(CGradientUpdateFunction *updateFunction); virtual ~CGradientFunctionUpdater() {}; virtual void updateWeights(CFeatureList *gradient) = 0; void addRandomParams(double randSize); CGradientUpdateFunction *getUpdateFunction() {return updateFunction;}; }; class CConstantGradientFunctionUpdater : public CGradientFunctionUpdater { public: CConstantGradientFunctionUpdater(CGradientUpdateFunction *updateFunction, double learningRate); virtual ~CConstantGradientFunctionUpdater() {}; virtual void updateWeights(CFeatureList *gradient); }; class CLineSearchGradientFunctionUpdater : public CGradientFunctionUpdater { protected: double *startParameters; double *workParameters; int maxSteps; CGradientCalculator *gradientCalculator; double precision_treshold; void setWorkingParamters(CFeatureList *gradient, double stepSize, double *startParameters, double *workParameters); virtual double getFunctionValue(double *startParameters, CFeatureList *gradient, double stepSize); void bracketMinimum(double *startParameters, CFeatureList *gradient, double fa, double &a, double &b, double &c); public: CLineSearchGradientFunctionUpdater(CGradientCalculator *gradientCalculator, CGradientUpdateFunction *updateFunction, int maxSteps); virtual ~CLineSearchGradientFunctionUpdater(); virtual void updateWeights(CFeatureList *gradient); virtual double updateWeights(CFeatureList *gradient, double fold, double &lmin); }; class CGradientLearner : virtual public CParameterObject { protected: CGradientCalculator *gradientCalculator; public: CGradientLearner(CGradientCalculator *gradientCalculator); virtual ~CGradientLearner() {}; virtual double doOptimization(int maxSteps) = 0; virtual void resetOptimization() {gradientCalculator->resetGradientCalculator();}; }; class CSupervisedGradientLearner : public CSupervisedLearner { protected: CGradientLearner *gradientLearner; CSupervisedGradientCalculator *gradientCalculator; public: CSupervisedGradientLearner(CGradientLearner *gradientLearner, CSupervisedGradientCalculator *gradientCalculator, int episodes); virtual ~CSupervisedGradientLearner(); virtual void learnFA(CDataSet *inputData, CDataSet1D *outputData); virtual void resetLearner(); }; class CSupervisedQFunctionLearnerFromLearners : public CSupervisedQFunctionLearner { protected: std::map<CAction *, CSupervisedLearner *> *learnerMap; public: CSupervisedQFunctionLearnerFromLearners(std::map<CAction *, CSupervisedLearner *> *learnerMap); virtual ~CSupervisedQFunctionLearnerFromLearners(); virtual void learnQFunction(CAction *action, CDataSet *inputData, CDataSet1D *outputData); virtual void resetLearner(); }; class CSupervisedQFunctionWeightedLearnerFromLearners : public CSupervisedQFunctionLearner { protected: std::map<CAction *, CSupervisedWeightedLearner *> *learnerMap; public: CSupervisedQFunctionWeightedLearnerFromLearners(std::map<CAction *, CSupervisedWeightedLearner *> *learnerMap); virtual ~CSupervisedQFunctionWeightedLearnerFromLearners(); virtual void learnQFunction(CAction *action, CDataSet *inputData, CDataSet1D *outputData, CDataSet1D *weightData); virtual void resetLearner(); }; class CBatchGradientLearner : public CGradientLearner { protected: CGradientFunctionUpdater *updater; CFeatureList *gradient; double treshold_f; public: CBatchGradientLearner(CGradientCalculator *gradientCalculator, CGradientFunctionUpdater *updater); ~CBatchGradientLearner(); virtual double doOptimization(int maxSteps); }; class CConjugateGradientLearner : public CGradientLearner { protected: CLineSearchGradientFunctionUpdater *gradientUpdater; CFeatureList *gradnew; CFeatureList *gradold; CFeatureList *d; double treshold_x; double treshold_f; double fnew; int exiting; public: CConjugateGradientLearner(CGradientCalculator *gradientCalculator, CLineSearchGradientFunctionUpdater *updater); virtual ~CConjugateGradientLearner(); virtual double doOptimization(int maxGradientUpdates); virtual void resetOptimization(); }; class CSupervisedNeuralNetworkMatlabLearner : public CSupervisedLearner { protected: CTorchGradientFunction *mlpFunction; public: CSupervisedNeuralNetworkMatlabLearner(CTorchGradientFunction *mlpFunction, int numHidden); virtual ~CSupervisedNeuralNetworkMatlabLearner(); virtual void learnFA(CDataSet *inputData, CDataSet1D *outputData); virtual void resetLearner(); }; class CSupervisedNeuralNetworkTorchLearner : public CSupervisedLearner, public CSupervisedWeightedLearner { protected: CTorchGradientFunction *mlpFunction; public: CSupervisedNeuralNetworkTorchLearner(CTorchGradientFunction *mlpFunction); virtual ~CSupervisedNeuralNetworkTorchLearner(); virtual void learnFA(CDataSet *inputData, CDataSet1D *outputData); virtual void learnWeightedFA(CDataSet *inputData, CDataSet1D *outputData, CDataSet1D *weighting); virtual void resetLearner(); }; #endif