My Neural Network Doesn't Work [XOR problem]

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)