tiny-dnn-xor.cpp

#include <iostream>
#include <vector>

#include "tiny_dnn/tiny_dnn.h"
using namespace tiny_dnn;
using namespace tiny_dnn::activation;
using namespace tiny_dnn::layers;

int main()
{
	network<sequential> net;
	net << fully_connected_layer(2,3) << sigmoid_layer()
		<< fully_connected_layer(3,1) << sigmoid_layer();

	std::vector<vec_t> trainIn = {{0,0}, {0,1}, {1,0}, {1,1}};
	std::vector<vec_t> trainOut = {{0}, {1}, {1}, {0}};

	gradient_descent optimizer(0.53f);
	net.fit<mse>(optimizer, trainIn, trainOut, 1, 1000);

	net.save("net");
	
	std::cout << net.predict({1,0})[0] << std::endl;
}

Good job but in this case you can also classify the output like this.

  network<sequential> net;
  net << fully_connected_layer(2,3) << sigmoid_layer()
      << fully_connected_layer(3,2) << sigmoid_layer();

  std::vector<vec_t> trainIn    = {{0,0}, {0,1}, {1,0}, {1,1}};
  std::vector<label_t> trainOut = {0, 1, 1, 0};

  gradient_descent optimizer;
  optimizer.alpha = 0.53f;
  timer t;
  t.start();
  net.train<cross_entropy>(optimizer, trainIn, trainOut, 1, 1000);
  t.stop();
  std::cout << "duration: " << t.total() << " s" << std::endl;

  net.save("net");

  std::cout << "result: " << net.predict_label({1,0}) << std::endl;
  std::cout << "probability: " << net.predict_max_value({1,0}) * 100 << " %" <<  std::endl;

duration: 51.0456 s
result: 1
probability: 99.5675 %