EIC code displayed by LXR |
|
|
|
|||
File indexing completed on 2025-01-18 10:11:11
0001 // @(#)root/tmva $Id$ 0002 // Author: Matt Jachowski 0003 0004 /********************************************************************************** 0005 * Project: TMVA - a Root-integrated toolkit for multivariate data analysis * 0006 * Package: TMVA * 0007 * Class : TMVA::TSynapse * 0008 * * 0009 * * 0010 * Description: * 0011 * Synapse class for use in derivatives of MethodANNBase * 0012 * * 0013 * Authors (alphabetical): * 0014 * Matt Jachowski <jachowski@stanford.edu> - Stanford University, USA * 0015 * * 0016 * Copyright (c) 2005: * 0017 * CERN, Switzerland * 0018 * * 0019 * Redistribution and use in source and binary forms, with or without * 0020 * modification, are permitted according to the terms listed in LICENSE * 0021 * (see tmva/doc/LICENSE) * 0022 **********************************************************************************/ 0023 0024 #ifndef ROOT_TMVA_TSynapse 0025 #define ROOT_TMVA_TSynapse 0026 0027 ////////////////////////////////////////////////////////////////////////// 0028 // // 0029 // TSynapse // 0030 // // 0031 // Synapse used by derivatives of MethodANNBase // 0032 // // 0033 ////////////////////////////////////////////////////////////////////////// 0034 0035 #include "TObject.h" 0036 0037 namespace TMVA { 0038 0039 class TNeuron; 0040 class MsgLogger; 0041 0042 class TSynapse : public TObject { 0043 0044 public: 0045 0046 TSynapse(); 0047 virtual ~TSynapse(); 0048 0049 // set the weight of the synapse 0050 void SetWeight(Double_t weight); 0051 0052 // get the weight of the synapse 0053 Double_t GetWeight() { return fWeight; } 0054 0055 // set the learning rate 0056 void SetLearningRate(Double_t rate) { fLearnRate = rate; } 0057 0058 // get the learning rate 0059 Double_t GetLearningRate() { return fLearnRate; } 0060 0061 // decay the learning rate 0062 void DecayLearningRate(Double_t rate){ fLearnRate *= (1-rate); } 0063 0064 // set the pre-neuron 0065 void SetPreNeuron(TNeuron* pre) { fPreNeuron = pre; } 0066 0067 // set the post-neuron 0068 void SetPostNeuron(TNeuron* post) { fPostNeuron = post; } 0069 0070 // get the weighted output of the pre-neuron 0071 Double_t GetWeightedValue(); 0072 0073 // get the weighted error field of the post-neuron 0074 Double_t GetWeightedDelta(); 0075 0076 // force the synapse to adjust its weight according to its error field 0077 void AdjustWeight(); 0078 0079 // calculate the error field of the synapse 0080 void CalculateDelta(); 0081 0082 // initialize the error field of the synapse to 0 0083 void InitDelta() { fDelta = 0.0; fCount = 0; } 0084 0085 void SetDEDw(Double_t DEDw) { fDEDw = DEDw; } 0086 Double_t GetDEDw() { return fDEDw; } 0087 Double_t GetDelta() { return fDelta; } 0088 0089 private: 0090 0091 Double_t fWeight; ///< weight of the synapse 0092 Double_t fLearnRate; ///< learning rate parameter 0093 Double_t fDelta; ///< local error field 0094 Double_t fDEDw; ///< sum of deltas 0095 Int_t fCount; ///< number of updates contributing to error field 0096 TNeuron* fPreNeuron; ///< pointer to pre-neuron 0097 TNeuron* fPostNeuron; ///< pointer to post-neuron 0098 0099 MsgLogger& Log() const; 0100 0101 ClassDef(TSynapse,0); // Synapse class used by MethodANNBase and derivatives 0102 }; 0103 0104 } // namespace TMVA 0105 0106 #endif
| [ Source navigation ] | [ Diff markup ] | [ Identifier search ] | [ general search ] |
|
This page was automatically generated by the 2.3.7 LXR engine. The LXR team |
|