int main()
{
string filename= "iris.data.txt";
vector<iovector> TrainSetIn; // 训练特征集合
vector<iovector> TrainSetOut;// 训练标注集合
vector<iovector> TestSetIn; // 测试集合
vector<iovector> TestSetOut;
if(0 != (Load(filename,TrainSetIn,TrainSetOut,TestSetIn,TestSetOut)))
{
return 1;
}
/* 初始化网络 */
CNeuralNet mynet( 4, 3, 10, 1);// 输入层维度，输出层维度，隐层维度，隐层层数
/* 训练网络 */
for(int epoch = 0; epoch < 1000; ++epoch )
mynet.TrainEpoch(TrainSetIn, TrainSetOut, 0.4);
/* 测试 */
double accuracy = 0;
int positive = 0;
for(int i = 0; i < TestSetIn.size(); ++i )
{
vector<double> output;
if(mynet.CalculateOutput(TestSetIn[i],output))
{
string TestLabel;
string TrueLabel;
GetLabel(TestSetOut[i],TrueLabel);
GetLabel(output,TestLabel);
cout<<"输出类别： "<<TestLabel <<" ("<<TrueLabel<<")" << endl;
if(0 ==(strcmp(TestLabel.c_str(), TrueLabel.c_str())))
positive ++;
}
}
accuracy = 1.0*positive / TestSetIn.size();
cout << "分类正确率：accuracy = " << accuracy*100 << "%" <<endl;
system("pause");
return 0;
}
int main(void) {
std::vector<std::string> datasets = std::vector<std::string>();
datasets.push_back("util/DS_AND.txt");
datasets.push_back("util/DS_OR.txt");
datasets.push_back("util/DS_NAND.txt");
datasets.push_back("util/DS_NOR.txt");
datasets.push_back("util/DS_XOR.txt");
datasets.push_back("util/DS_XNOR.txt");
for(std::string ds : datasets) {
std::cout << ds << std::endl;
TrainingData td(ds);
std::vector<unsigned> topo;
td.getTopology(topo);
Neural::Network mynet(topo);
std::vector<double> inputs, targets;//, outputs;
unsigned trainingPass = 0;
while(!td.isEOF()) {
trainingPass++;
std::cout << trainingPass << ";";
if(td.getNextInputs(inputs) != topo[0]) {
std::cout << -1 << std::endl;
break;
}
mynet.feedForward(inputs);
// mynet.getResults(outputs);
td.getNextTargets(targets);
assert(targets.size()==topo.back());
mynet.backProp(targets);
std::cout << mynet.getRecentAvgError() << std::endl;
}
}
/*
TrainingData td("trainingData.txt");
std::vector<unsigned> topo;
td.getTopology(topo);
Neural::Network mynet(topo);
std::vector<double> inputs, targets, outputs;
unsigned trainingPass = 0;
while(!td.isEOF()) {
trainingPass++;
std::cout << std::endl << "Pass " << trainingPass;
if(td.getNextInputs(inputs) != topo[0])
break;
printVector(": Inputs", inputs);
mynet.feedForward(inputs);
mynet.getResults(outputs);
printVector("Outputs", outputs);
td.getNextTargets(targets);
printVector("Targets", targets);
assert(targets.size()==topo.back());
mynet.backProp(targets);
std::cout << "Net recent avg error: " << mynet.getRecentAvgError() << std::endl;
}
std::cout << std::endl << "Done" << std::endl;
*/
return 0;
}