net_test_2.cpp

/**
 * Example of a neural network with reused edge weights
 */

//
// Created by Michal on 7/17/18.
//

#include <vector>

#include "4neuro.h"

int main() {
    std::cout << "Running lib4neuro example   2: Basic use of the particle swarm method to train a network with five linear neurons and repeating edge weights" << std::endl;
    std::cout << "********************************************************************************************************************************************" <<std::endl;
    std::cout << "The code attempts to find an approximate solution to the system of equations below:" << std::endl;
    std::cout << " 0 * w1 + 1 * w2 = 0.50 + b1" << std::endl;
    std::cout << " 1 * w1 + 0.5*w2 = 0.75 + b1" << std::endl;
    std::cout << "(1.25 + b2) * w2 = 0.63 + b3" << std::endl;
    std::cout << "***********************************************************************************************************************" <<std::endl;

    /* TRAIN DATA DEFINITION */
    std::vector<std::pair<std::vector<double>, std::vector<double>>> data_vec;
    std::vector<double> inp, out;

    inp = {0, 1, 0};
    out = {0.5, 0};
    data_vec.emplace_back(std::make_pair(inp, out));

    inp = {1, 0.5, 0};
    out = {0.75, 0};
    data_vec.emplace_back(std::make_pair(inp, out));

    inp = {0, 0, 1.25};
    out = {0, 0.63};
    data_vec.emplace_back(std::make_pair(inp, out));
    l4n::DataSet ds(&data_vec);

    /* NETWORK DEFINITION */
    l4n::NeuralNetwork net;

    /* Input neurons */
    l4n::NeuronLinear *i1 = new l4n::NeuronLinear( );  //f(x) = x
    l4n::NeuronLinear *i2 = new l4n::NeuronLinear( );  //f(x) = x

    double b = 1;//bias
    l4n::NeuronLinear *i3 = new l4n::NeuronLinear( );  //f(x) = x

    /* Output neurons */
    l4n::NeuronLinear *o1 = new l4n::NeuronLinear( );  //f(x) = x
    l4n::NeuronLinear *o2 = new l4n::NeuronLinear( );  //f(x) = x



    /* Adding neurons to the nets */
    size_t idx1 = net.add_neuron(i1, l4n::BIAS_TYPE::NO_BIAS);
    size_t idx2 = net.add_neuron(i2, l4n::BIAS_TYPE::NO_BIAS);
    size_t idx3 = net.add_neuron(o1, l4n::BIAS_TYPE::NEXT_BIAS);
    size_t idx4 = net.add_neuron(i3, l4n::BIAS_TYPE::NEXT_BIAS);
    size_t idx5 = net.add_neuron(o2, l4n::BIAS_TYPE::NEXT_BIAS);

    /* Adding connections */
    net.add_connection_simple(idx1, idx3, l4n::SIMPLE_CONNECTION_TYPE::NEXT_WEIGHT); // weight index 0
    net.add_connection_simple(idx2, idx3, l4n::SIMPLE_CONNECTION_TYPE::NEXT_WEIGHT); // weight index 1
    net.add_connection_simple(idx4, idx5, l4n::SIMPLE_CONNECTION_TYPE::EXISTING_WEIGHT, 0); // AGAIN weight index 0 - same weight!

    net.randomize_weights();

    /* specification of the input/output neurons */
    std::vector<size_t> net_input_neurons_indices(3);
    std::vector<size_t> net_output_neurons_indices(2);
    net_input_neurons_indices[0] = idx1;
    net_input_neurons_indices[1] = idx2;
    net_input_neurons_indices[2] = idx4;

    net_output_neurons_indices[0] = idx3;
    net_output_neurons_indices[1] = idx5;

    net.specify_input_neurons(net_input_neurons_indices);
    net.specify_output_neurons(net_output_neurons_indices);
    

    /* COMPLEX ERROR FUNCTION SPECIFICATION */
    l4n::MSE mse(&net, &ds);

//    double weights[2] = {-0.18012411, -0.17793740};
//    double weights[2] = {1, 1};

//    printf("evaluation of error at point (%f, %f) => %f\n", weights[0], weights[1], mse.eval(weights));

    /* TRAINING METHOD SETUP */
    //must encapsulate each of the partial error functions
    std::vector<double> domain_bounds = {-10.0, 10.0, -10.0, 10.0,-10.0, 10.0, -10.0, 10.0,-10.0, 10.0, -10.0, 10.0,-10.0, 10.0, -10.0, 10.0,-10.0, 10.0, -10.0, 10.0,-10.0, 10.0, -10.0, 10.0,-10.0, 10.0, -10.0, 10.0};
    ParticleSwarm swarm_01(&mse, &domain_bounds);

    /* if the maximal velocity from the previous step is less than 'gamma' times the current maximal velocity, then one
     * terminating criterion is met */
    double gamma = 0.5;

    /* if 'delta' times 'n' particles are in the centroid neighborhood given by the radius 'epsilon', then the second
     * terminating criterion is met ('n' is the total number of particles) */
    double epsilon = 0.02;
    double delta = 0.9;
    swarm_01.optimize(gamma, epsilon, delta);

    std::vector<double> *parameters = swarm_01.get_solution();
    net.copy_parameter_space(parameters);

    printf("w1 = %10.7f\n", parameters->at( 0 ));
    printf("w2 = %10.7f\n", parameters->at( 1 ));
    printf("b1 = %10.7f\n", parameters->at( 2 ));
    printf("b2 = %10.7f\n", parameters->at( 3 ));
    printf("b3 = %10.7f\n", parameters->at( 4 ));
    std::cout << "***********************************************************************************************************************" <<std::endl;

    /* ERROR CALCULATION */
    double error = 0.0;
    inp = {0, 1, 0};
    net.eval_single( inp, out );
    error += (0.5 - out[0]) * (0.5 - out[0]) + (0.0 - out[1]) * (0.0 - out[1]);
    printf("x = (%4.2f, %4.2f, %4.2f), expected output: (%4.2f, %4.2f), real output: (%10.7f, %10.7f)\n", inp[0], inp[1], inp[2], 0.5, 0.0, out[0], out[1]);

    inp = {1, 0.5, 0};
    net.eval_single( inp, out );
    error += (0.75 - out[0]) * (0.75 - out[0]) + (0.0 - out[1]) * (0.0 - out[1]);
    printf("x = (%4.2f, %4.2f, %4.2f), expected output: (%4.2f, %4.2f), real output: (%10.7f, %10.7f)\n", inp[0], inp[1], inp[2], 0.75, 0.0, out[0], out[1]);

    inp = {0, 0, 1.25};
    net.eval_single( inp, out );
    error += (0.0 - out[0]) * (0.0 - out[0]) + (0.63 - out[1]) * (0.63 - out[1]);
    printf("x = (%4.2f, %4.2f, %4.2f), expected output: (%4.2f, %4.2f), real output: (%10.7f, %10.7f)\n", inp[0], inp[1], inp[2], 0.0, 0.63, out[0], out[1]);

    std::cout << "Run finished! Error of the network: " << error / 3.0 << std::endl;



    return 0;
}