#include "../include/Network.h" Network::Network(Genome &g) { for (int i = 0; i < g.inputs; ++i) { inputs.push_back(new Neuron(i)); } for (int i = 0; i < g.outputs; ++i) { outputs.push_back(new Neuron(g.inputs + i)); } for (int i = 0; i < g.structure.size(); ++i) { //Only constructs enabled connections for performance if (g.structure[i].enabled) { //Computes Inputs structure.push_back(Connection(g.structure[i].weight)); if (g.structure[i].input < g.inputs) { structure.back().in = inputs[g.structure[i].input]; } else if (g.structure[i].input < g.inputs + g.outputs) { structure.back().in = inputs[g.structure[i].input - g.inputs]; } else { bool temp = true; for (int j = 0; j < hiddenNodes.size(); ++j) { if (hiddenNodes[j]->id == g.structure[i].input) { temp = false; structure.back().in = hiddenNodes[j]; } } if (temp) { hiddenNodes.push_back(new Neuron(g.structure[i].input)); structure.back().in = hiddenNodes.back(); } } //Computes Outputs if (g.structure[i].output < g.inputs + g.outputs) { structure.back().out = outputs[g.structure[i].output - g.inputs]; } else { bool temp = true; for (int j = 0; j < hiddenNodes.size(); ++j) { if (hiddenNodes[j]->id == g.structure[i].output) { temp = false; structure.back().out = hiddenNodes[j]; } } if (temp) { hiddenNodes.push_back(new Neuron(g.structure[i].output)); structure.back().out = hiddenNodes.back(); } } } } } double Network::sigmoid(double a) { return 2. / (1 + exp(-4.9 * a)) - 1; } void Network::propagate(double a[], double b[]) { for (int i = 0; i < inputs.size(); ++i) { inputs[i]->value = a[i]; } for (int i = 0; i < hiddenNodes.size(); ++i) { hiddenNodes[i]->sum = 0; } for (int i = 0; i < outputs.size(); ++i) { outputs[i]->sum = 0; } for (int i = 0; i < structure.size(); ++i) { structure[i].out->sum += structure[i].in->value * structure[i].weight; //Recomputes values in real time. might be a performance hit //solution would be to do it as a batch after propagation //that would take more steps to accurately propagate though structure[i].out->value = sigmoid(structure[i].out->sum); } for (int i = 0; i < outputs.size(); ++i) { b[i] = outputs[i]->value; } } void Network::reset() { for (int i = 0; i < inputs.size(); ++i) { inputs[i]->value = 0; } for (int i = 0; i < hiddenNodes.size(); ++i) { hiddenNodes[i]->value = 0; } for (int i = 0; i < outputs.size(); ++i) { outputs[i]->value = 0; } } Network::~Network() { for (int i = 0; i < inputs.size(); ++i) { delete inputs[i]; } for (int i = 0; i < hiddenNodes.size(); ++i) { delete hiddenNodes[i]; } for (int i = 0; i < outputs.size(); ++i) { delete outputs[i]; } }