本文整理汇总了C++中NeuralNetwork::computeCost方法的典型用法代码示例。如果您正苦于以下问题：C++ NeuralNetwork::computeCost方法的具体用法？C++ NeuralNetwork::computeCost怎么用？C++ NeuralNetwork::computeCost使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类NeuralNetwork的用法示例。
在下文中一共展示了NeuralNetwork::computeCost方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的C++代码示例。
示例1: main
int main(int argc, char* argv[])
{
cout << "Reading in the input data" << endl;
/************** Read in images *****************/
int imgSize = 512;
int numImages = 10;
string filename = "olsh.dat";
vector<MatrixXd> images;
ifstream indata;
double num; // variable for input value
indata.open(filename.c_str());
if(!indata) {
cerr << "Error: file could not be opened" << endl;
return 1;
}
for(int i = 0; i < numImages; ++i) {
MatrixXd m(imgSize,imgSize);
for(int r = 0; r < imgSize; ++r) {
for(int c = 0; c < imgSize; ++c) {
if(indata.eof()) {
cerr << "Error: ran out of input values on (" << r << "," << c << ")" << endl;
return 1;
}
indata >> num;
m(r,c) = num;
}
}
images.push_back(m);
}
indata.close();
cout << "Input data loaded" << endl;
/*************** YOUR CODE HERE ************/
/* parse the command line options */
oFilename = NULL;
if (argc > 1) {
sscanf(argv[1], "%lf", &alpha);
}
if (argc > 2) {
sscanf(argv[2], "%d", &numIterations);
}
if (argc > 3) {
sscanf(argv[3], "%lf", &beta);
}
if (argc > 4) {
oFilename = argv[4];
cout << "Outputting to file " << oFilename << endl;
}
else {
oFilename = (char *)malloc(256);
sprintf(oFilename, "bases/alpha%fnumIterations%d.dat", alpha, numIterations);
}
/* For each training example:
(i) Run a forward pass on our network on input x, to compute all
units? activations;
(ii) Perform one step of stochas- tic gradient descent using
backpropagation;
(iii) Perform the updates given in Equations (8-9).
*/
int unitCounts[] = {WINDOW_PIXEL_SIZE * WINDOW_PIXEL_SIZE,
30,
WINDOW_PIXEL_SIZE * WINDOW_PIXEL_SIZE};
NeuralNetwork nn = NeuralNetwork(3, unitCounts);
MatrixXd trainingImage;
cout << "Performing " << numIterations << " iterations with alpha = " << alpha << endl;
for (int iteration=0; iteration < numIterations; iteration++) {
generateTrainingExample(images, trainingImage);
if (iteration % 1000 == 0 || iteration < 11) {
cout << iteration << " " << flush;
cout << endl << " ";
nn.computeCost(trainingImage, trainingImage);
}
// (i) run a feedforward pass on our network
nn.forwardPropagateActivations(trainingImage);
// (ii) perform one step of stochastic gradient descent
nn.backprop(trainingImage);
// (iii) perform the updates given in Equations (8-9).
nn.sparseLearningPass(iteration % 1000 == 0 || iteration < 11);
if (iteration % 1000 == 0 || iteration < 11) {
cout << " ";
nn.computeCost(trainingImage, trainingImage);
}
}
//.........这里部分代码省略.........