My Neural Network Doesn't Work [XOR problem] - c++

Question

I'm trying to make a neural network for solving XOR problem.But I couldn't make it.Always giving false results.Maybe I'm making a mistake in your math.The network does not learn.Result always similarly.
I am not using BIAS.
Note: execute function = (feed-forward + backpropagation)
ALPHA = 0.5
Here is the code:
//main.cpp
#include <iostream>
#include "neural_network.h"
int main(int argc, char const *argv[])
{
srand(time(NULL));
double array[][3] = {{0.0, 0.0, 0.0},
{0.0, 1.0, 1.0},
{1.0, 0.0, 1.0},
{1.0, 1.0, 0.0}};
neural_network* nn = new neural_network(3, 2, 2, 1, 1.0);
nn->create_network();
for(int i = 0; i < 15000; i++)
{
int index = rand() % 4;
#if DEBUG
std::cout<<"Inputs :"<<array[index][0]<<" , "<<array[index][1]<<std::endl;
std::cout<<"Outputs :"<<array[index][2]<<std::endl;
#endif
nn->execute(array[index], &array[index][2]);
}
nn->print_weight();
nn->execute(array[0], &array[0][2]);
nn->print_output();
nn->execute(array[1], &array[1][2]);
nn->print_output();
nn->execute(array[2], &array[2][2]);
nn->print_output();
nn->execute(array[3], &array[3][2]);
nn->print_output();
return 0;
}
//feed-forward function
void neural_network::feed_forward(double* inputs)
{
int index = 0;
for(int i = 0; i < neural_network::input_layer_size; i++)
neural_network::input_neuron[i] = inputs[i];
for(int i = 0; i < neural_network::hidden_layer_size; i++)
{
for(int j = 0; j < neural_network::input_layer_size; j++)
{
neural_network::hidden_neuron[i] += neural_network::input_neuron[j] * weight_I_H[index++];
}
neural_network::hidden_neuron[i] = neural_network::activation_func(neural_network::hidden_neuron[i]);
}
index = 0;
for(int i = 0; i < neural_network::output_layer_size; i++)
{
for(int j = 0; j < neural_network::hidden_layer_size; j++)
{
neural_network::output_neuron[i] += neural_network::hidden_neuron[j] * weight_H_O[index++];
}
neural_network::output_neuron[i] = neural_network::activation_func(neural_network::output_neuron[i]);
}
}
//backpropagation function
void neural_network::back_propagation(double* outputs)
{
int index;
for(int i = 0; i < neural_network::output_layer_size; i++)
neural_network::err_output[i] = (outputs[i] - neural_network::output_neuron[i]);
for(int i = 0; i < neural_network::hidden_layer_size; i++)
{
index = i;
for(int j = 0; j < neural_network::output_layer_size; j++)
{
neural_network::err_hidden[i] += neural_network::weight_H_O[index] * neural_network::err_output[j] * neural_network::derivative_act_func(neural_network::output_neuron[j]);
neural_network::weight_H_O[index] += ALPHA * neural_network::err_output[j] * neural_network::derivative_act_func(neural_network::output_neuron[j]) * neural_network::hidden_neuron[i];
index += neural_network::hidden_layer_size;
}
}
for(int i = 0; i < neural_network::input_layer_size; i++)
{
index = i;
for(int j = 0; j < neural_network::hidden_layer_size; j++)
{
neural_network::weight_I_H[index] += ALPHA * neural_network::err_hidden[j] * neural_network::derivative_act_func(neural_network::hidden_neuron[j]) * neural_network::input_neuron[i];
index += neural_network::input_layer_size;
}
}
}
//output
Input To Hidden :
H-1 :
Weight :-13.269
Weight :-13.2705
H-2 :
Weight :-12.5172
Weight :-12.5195
Hidden To Output :
O-1 :
Weight :-5.37707
Weight :-2.93218
Outputs for (0,0):
O-1 :0.0294265
Outputs for (0,1):
O-1 :0.507348
Outputs for (1,0):
O-1 :0.62418
Outputs for (1,1):
O-1 :0.651169

Answer 1

It is real impossible no keras no my developed net based on Furye transformation(which is more power than keras) real decide this XOR task.I tested very accuracy both of this ANN.Maximum the recognize is 3 examples of 4 (acc=0.75->75%).No one answered 1 xor 1=0.It seems nowbody realy tested this case seriosly.(ANN were multilayered)