//training data:nDim * nExamples

#include "dataAndImage.h"

#include "SAE.h"

#include "cmath"

#include "getConfig.h"

int main()

{

//regularization coefficient

double lambda = 0.0001;

//learning rate

double alpha = 0.03;

double beta = 3;

double sp = 0.01;

int maxIter = 15000;

int miniBatchSize = 1000;

int n = 2;

int inputSize = 28*28;

int hiddenSize = 28;

int imgWidth = 8;

char *fileBuf = new char[4096];

bool ret = loadFileToBuf("ParamConfig.ini",fileBuf,4096);

if(ret)

{

getConfigDoubleValue(fileBuf,"lambda:",lambda);

getConfigDoubleValue(fileBuf,"alpha:",alpha);

getConfigDoubleValue(fileBuf,"beta:",beta);

getConfigDoubleValue(fileBuf,"sparseParam:",sp);

getConfigIntValue(fileBuf,"maxIter:",maxIter);

getConfigIntValue(fileBuf,"miniBatchSize:",miniBatchSize);

getConfigIntValue(fileBuf,"hiddenSize:",hiddenSize);

getConfigIntValue(fileBuf,"inputSize:",inputSize);

getConfigIntValue(fileBuf,"imgWidth:",imgWidth);

cout << "lambda:" << lambda << endl;

cout << "alpha:" << alpha << endl;

cout << "beta:" << beta << endl;

cout << "sparseParam:" << sp << endl;

cout << "maxIter:" << maxIter << endl;

cout << "miniBatchSize:" << miniBatchSize << endl;

cout << "hiddenSize:" << hiddenSize << endl;

cout << "inputSize:" << inputSize << endl;

cout << "imgWidth:" << imgWidth << endl;

}

delete []fileBuf;

MatrixXd trainData(n,inputSize);

SAE sae(inputSize,hiddenSize);

ret = loadMnistData(trainData,"mnist\\train-images-idx3-ubyte");

cout << "Loading training data..." << endl;

if(ret == false)

{

return -1;

}

//buildImage(trainData.topRows(100),8,"data.bmp");

//sae.loadModel("SAE_Model.txt");

clock_t start = clock();

//cout << trainData.rows() << " " << trainData.cols() << endl;

MatrixXd showImage = trainData.leftCols(100).transpose();

buildImage(showImage,imgWidth,"data.jpg");

sae.train(trainData,lambda,alpha,beta,sp,maxIter,MINI_BATCH_SGD,&miniBatchSize);

cout << "End Train" << endl;

MatrixXd hiddenTheta = sae.getTheta();

buildImage(hiddenTheta,imgWidth,"weights.jpg",true);

cout << "Saving hidden neurons" << endl;

sae.saveModel("SAE_Model.txt");

clock_t end = clock();

cout << "The code ran for " << (end - start)/(double)(CLOCKS_PER_SEC*60) << " minutes." << endl;

/*sae.loadModel("SAE_Model.txt");

buildImage(sae.theta1,imgWidth,"weights.bmp",true);*/

cout << "lambda:" << lambda << endl;

cout << "alpha:" << alpha << endl;

cout << "beta:" << beta << endl;

cout << "sparseParam:" << sp << endl;

cout << "miniBatchSize:" << miniBatchSize << endl;

system("pause");

return 0;

}