From 2c525d9391ef076ec440b070bd8678f3b3f81808 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?C=C3=A9dric=20Traizet?= <traizetc@cesbio.cnes.fr>
Date: Tue, 27 Jun 2017 14:50:39 +0200
Subject: [PATCH] Training an autoencoder now outputs a learning curve (txt
file), work in progress
---
include/AutoencoderModel.h | 10 ++++++----
include/AutoencoderModel.txx | 21 +++++++++++++++------
2 files changed, 21 insertions(+), 10 deletions(-)
diff --git a/include/AutoencoderModel.h b/include/AutoencoderModel.h
index 72dd370568..c6038427d2 100644
--- a/include/AutoencoderModel.h
+++ b/include/AutoencoderModel.h
@@ -3,7 +3,7 @@
#include "otbMachineLearningModelTraits.h"
#include "otbMachineLearningModel.h"
-
+#include <fstream>
#include <shark/Algorithms/StoppingCriteria/AbstractStoppingCriterion.h>
namespace otb
@@ -68,10 +68,10 @@ public:
void Train() ITK_OVERRIDE;
template <class T>
- void TrainOneLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int,double, double, shark::Data<shark::RealVector> &);
+ void TrainOneLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int,double, double, shark::Data<shark::RealVector> &, std::ostream&);
template <class T>
- void TrainOneSparseLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int,double, double,double, shark::Data<shark::RealVector> &);
+ void TrainOneSparseLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int,double, double,double, shark::Data<shark::RealVector> &, std::ostream&);
protected:
AutoencoderModel();
@@ -85,7 +85,9 @@ private:
/** Network attributes */
std::vector<AutoencoderType> m_net;
-
+
+ typename InputListSampleType::Pointer m_ValidationListSample; // This list can optionnaly be used during training to prevent overfitting.
+
itk::Array<unsigned int> m_NumberOfHiddenNeurons;
/** Training parameters */
diff --git a/include/AutoencoderModel.txx b/include/AutoencoderModel.txx
index fd71a73452..ecd39a468f 100644
--- a/include/AutoencoderModel.txx
+++ b/include/AutoencoderModel.txx
@@ -43,6 +43,9 @@ void AutoencoderModel<TInputValue,AutoencoderType>::Train()
shark::Data<shark::RealVector> inputSamples = shark::createDataFromRange( features );
+ std::ofstream ofs("/mnt/data/home/traizetc/computation/learning_curve.txt"); //learning curve
+ ofs << "learning curve" << std::endl;
+
if (m_Epsilon > 0){
shark::TrainingProgress<> criterion(5,m_Epsilon);
@@ -50,11 +53,11 @@ void AutoencoderModel<TInputValue,AutoencoderType>::Train()
{
if (m_Noise[i] != 0) // Shark doesn't allow to train a layer using a sparsity term AND a noisy input. (shark::SparseAutoencoderError takes an autoen
{
- TrainOneLayer(criterion, m_NumberOfHiddenNeurons[i],m_Noise[i],m_Regularization[i], inputSamples);
+ TrainOneLayer(criterion, m_NumberOfHiddenNeurons[i],m_Noise[i],m_Regularization[i], inputSamples,ofs);
}
else
{
- TrainOneSparseLayer( criterion,m_NumberOfHiddenNeurons[i],m_Rho[i],m_Beta[i],m_Regularization[i], inputSamples);
+ TrainOneSparseLayer( criterion,m_NumberOfHiddenNeurons[i],m_Rho[i],m_Beta[i],m_Regularization[i],inputSamples, ofs);
}
criterion.reset();
}
@@ -68,11 +71,11 @@ void AutoencoderModel<TInputValue,AutoencoderType>::Train()
{
if (m_Noise[i] != 0) // Shark doesn't allow to train a layer using a sparsity term AND a noisy input. (shark::SparseAutoencoderError takes an autoen
{
- TrainOneLayer(criterion, m_NumberOfHiddenNeurons[i],m_Noise[i],m_Regularization[i], inputSamples);
+ TrainOneLayer(criterion, m_NumberOfHiddenNeurons[i],m_Noise[i],m_Regularization[i], inputSamples, ofs);
}
else
{
- TrainOneSparseLayer(criterion, m_NumberOfHiddenNeurons[i],m_Rho[i],m_Beta[i],m_Regularization[i], inputSamples);
+ TrainOneSparseLayer(criterion, m_NumberOfHiddenNeurons[i],m_Rho[i],m_Beta[i],m_Regularization[i], inputSamples, ofs);
}
criterion.reset();
}
@@ -82,7 +85,7 @@ void AutoencoderModel<TInputValue,AutoencoderType>::Train()
template <class TInputValue, class AutoencoderType>
template <class T>
-void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int nbneuron,double noise_strength,double regularization, shark::Data<shark::RealVector> &samples)
+void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int nbneuron,double noise_strength,double regularization, shark::Data<shark::RealVector> &samples, std::ostream& File)
{
AutoencoderType net;
@@ -103,10 +106,15 @@ void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneLayer(shark::Abstrac
optimizer.init(error);
std::cout<<"error before training : " << optimizer.solution().value<<std::endl;
+
+ File << "end layer" << std::endl;
+
+
unsigned int i=0;
do{
i++;
optimizer.step(error);
+ File << optimizer.solution().value << std::endl;
} while( !criterion.stop( optimizer.solution() ) );
std::cout<<"error after " << i << "iterations : " << optimizer.solution().value<<std::endl;
@@ -118,7 +126,7 @@ void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneLayer(shark::Abstrac
template <class TInputValue, class AutoencoderType>
template <class T>
-void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneSparseLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int nbneuron,double rho,double beta, double regularization, shark::Data<shark::RealVector> &samples)
+void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneSparseLayer(shark::AbstractStoppingCriterion<T> & criterion, unsigned int nbneuron,double rho,double beta, double regularization, shark::Data<shark::RealVector> &samples, std::ostream& File)
{
AutoencoderType net;
@@ -141,6 +149,7 @@ void AutoencoderModel<TInputValue,AutoencoderType>::TrainOneSparseLayer(shark::A
do{
i++;
optimizer.step(error);
+ File << optimizer.solution().value << std::endl;
} while( !criterion.stop( optimizer.solution() ) );
std::cout<<"error after " << i << "iterations : " << optimizer.solution().value<<std::endl;
--
GitLab