Relintai/pmlpp
1
0
Fork 0
mirror of https://github.com/Relintai/pmlpp.git synced 2024-12-22 15:06:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

Now MLPPGAN uses engine classes.

Browse Source
This commit is contained in:
Relintai 2023-02-16 17:32:35 +01:00
parent 3c8ee1ffea
commit 737b34f53d
2 changed files with 183 additions and 147 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

283
mlpp/gan/gan.cpp
View File

@ -11,6 +11,8 @@
#include "../regularization/reg.h" #include "../regularization/reg.h"
#include "../utilities/utilities.h" #include "../utilities/utilities.h"
#include "core/log/logger.h"
#include <cmath> #include <cmath>
#include <iostream> #include <iostream>
@ -37,10 +39,10 @@ void MLPPGAN::set_k(const int val) {
} }
*/ */
std::vector<std::vector<real_t>> MLPPGAN::generate_example(int n) { Ref<MLPPMatrix> MLPPGAN::generate_example(int n) {
MLPPLinAlg alg; MLPPLinAlg alg;
return model_set_test_generator(alg.gaussianNoise(n, _k)); return model_set_test_generator(alg.gaussian_noise(n, _k));
} }
void MLPPGAN::gradient_descent(real_t learning_rate, int max_epoch, bool ui) { void MLPPGAN::gradient_descent(real_t learning_rate, int max_epoch, bool ui) {
@ -52,41 +54,39 @@ void MLPPGAN::gradient_descent(real_t learning_rate, int max_epoch, bool ui) {
forward_pass(); forward_pass();
while (true) { while (true) {
cost_prev = cost(_y_hat, alg.onevec(_n)); cost_prev = cost(_y_hat, alg.onevecv(_n));
// Training of the discriminator. // Training of the discriminator.
std::vector<std::vector<real_t>> generator_input_set = alg.gaussianNoise(_n, _k); Ref<MLPPMatrix> generator_input_set = alg.gaussian_noise(_n, _k);
std::vector<std::vector<real_t>> discriminator_input_set = model_set_test_generator(generator_input_set); Ref<MLPPMatrix> discriminator_input_set = model_set_test_generator(generator_input_set);
discriminator_input_set.insert(discriminator_input_set.end(), _output_set.begin(), _output_set.end()); // Fake + real inputs. discriminator_input_set->add_rows_mlpp_matrix(_output_set); // Fake + real inputs.
std::vector<real_t> y_hat = model_set_test_discriminator(discriminator_input_set); Ref<MLPPVector> y_hat = model_set_test_discriminator(discriminator_input_set);
std::vector<real_t> _output_set = alg.zerovec(_n); Ref<MLPPVector> output_set = alg.zerovecv(_n);
std::vector<real_t> _output_setReal = alg.onevec(_n); Ref<MLPPVector> output_set_real = alg.onevecv(_n);
_output_set.insert(_output_set.end(), _output_setReal.begin(), _output_setReal.end()); // Fake + real output scores. output_set->add_mlpp_vector(output_set_real); // Fake + real output scores.
auto dgrads = compute_discriminator_gradients(y_hat, _output_set); ComputeDiscriminatorGradientsResult dgrads = compute_discriminator_gradients(y_hat, _output_set);
auto cumulative_discriminator_hidden_layer_w_grad = std::get<0>(dgrads);
auto outputDiscriminatorWGrad = std::get<1>(dgrads);
cumulative_discriminator_hidden_layer_w_grad = alg.scalarMultiply(learning_rate / _n, cumulative_discriminator_hidden_layer_w_grad); dgrads.cumulative_hidden_layer_w_grad = alg.scalar_multiply_vm(learning_rate / _n, dgrads.cumulative_hidden_layer_w_grad);
outputDiscriminatorWGrad = alg.scalarMultiply(learning_rate / _n, outputDiscriminatorWGrad); dgrads.output_w_grad = alg.scalar_multiplynv(learning_rate / _n, dgrads.output_w_grad);
update_discriminator_parameters(cumulative_discriminator_hidden_layer_w_grad, outputDiscriminatorWGrad, learning_rate); update_discriminator_parameters(dgrads.cumulative_hidden_layer_w_grad, dgrads.output_w_grad, learning_rate);
// Training of the generator. // Training of the generator.
generator_input_set = alg.gaussianNoise(_n, _k); generator_input_set = alg.gaussian_noise(_n, _k);
discriminator_input_set = model_set_test_generator(generator_input_set); discriminator_input_set = model_set_test_generator(generator_input_set);
y_hat = model_set_test_discriminator(discriminator_input_set); y_hat = model_set_test_discriminator(discriminator_input_set);
_output_set = alg.onevec(_n); _output_set = alg.onevecv(_n);
std::vector<std::vector<std::vector<real_t>>> cumulative_generator_hidden_layer_w_grad = compute_generator_gradients(y_hat, _output_set); Vector<Ref<MLPPMatrix>> cumulative_generator_hidden_layer_w_grad = compute_generator_gradients(y_hat, _output_set);
cumulative_generator_hidden_layer_w_grad = alg.scalarMultiply(learning_rate / _n, cumulative_generator_hidden_layer_w_grad); cumulative_generator_hidden_layer_w_grad = alg.scalar_multiply_vm(learning_rate / _n, cumulative_generator_hidden_layer_w_grad);
update_generator_parameters(cumulative_generator_hidden_layer_w_grad, learning_rate); update_generator_parameters(cumulative_generator_hidden_layer_w_grad, learning_rate);
forward_pass(); forward_pass();
if (ui) { if (ui) {
print_ui(epoch, cost_prev, _y_hat, alg.onevec(_n)); print_ui(epoch, cost_prev, _y_hat, alg.onevecv(_n));
} }
epoch++; epoch++;
@ -103,46 +103,54 @@ real_t MLPPGAN::score() {
forward_pass(); forward_pass();
return util.performance(_y_hat, alg.onevec(_n)); return util.performance_vec(_y_hat, alg.onevecv(_n));
} }
void MLPPGAN::save(std::string fileName) { void MLPPGAN::save(const String &file_name) {
MLPPUtilities util; MLPPUtilities util;
/*
if (!_network.empty()) { if (!_network.empty()) {
util.saveParameters(fileName, _network[0].weights, _network[0].bias, false, 1); util.saveParameters(file_name, _network[0].weights, _network[0].bias, false, 1);
for (uint32_t i = 1; i < _network.size(); i++) { for (uint32_t i = 1; i < _network.size(); i++) {
util.saveParameters(fileName, _network[i].weights, _network[i].bias, true, i + 1); util.saveParameters(file_name, _network[i].weights, _network[i].bias, true, i + 1);
} }
util.saveParameters(fileName, _output_layer->weights, _output_layer->bias, true, _network.size() + 1); util.saveParameters(file_name, _output_layer->weights, _output_layer->bias, true, _network.size() + 1);
} else { } else {
util.saveParameters(fileName, _output_layer->weights, _output_layer->bias, false, _network.size() + 1); util.saveParameters(file_name, _output_layer->weights, _output_layer->bias, false, _network.size() + 1);
} }
*/
} }
void MLPPGAN::add_layer(int n_hidden, std::string activation, std::string weight_init, std::string reg, real_t lambda, real_t alpha) { void MLPPGAN::add_layer(int n_hidden, MLPPActivation::ActivationFunction activation, MLPPUtilities::WeightDistributionType weight_init, MLPPReg::RegularizationType reg, real_t lambda, real_t alpha) {
MLPPLinAlg alg; MLPPLinAlg alg;
if (_network.empty()) { if (_network.empty()) {
_network.push_back(MLPPOldHiddenLayer(n_hidden, activation, alg.gaussianNoise(_n, _k), weight_init, reg, lambda, alpha)); Ref<MLPPHiddenLayer> layer = Ref<MLPPHiddenLayer>(memnew(MLPPHiddenLayer(n_hidden, activation, alg.gaussian_noise(_n, _k), weight_init, reg, lambda, alpha)));
_network[0].forwardPass();
_network.push_back(layer);
_network.write[0]->forward_pass();
} else { } else {
_network.push_back(MLPPOldHiddenLayer(n_hidden, activation, _network[_network.size() - 1].a, weight_init, reg, lambda, alpha)); Ref<MLPPHiddenLayer> layer = Ref<MLPPHiddenLayer>(memnew(MLPPHiddenLayer(n_hidden, activation, _network.write[_network.size() - 1]->get_a(), weight_init, reg, lambda, alpha)));
_network[_network.size() - 1].forwardPass();
_network.push_back(layer);
_network.write[_network.size() - 1]->forward_pass();
} }
} }
void MLPPGAN::add_output_layer(std::string weight_init, std::string reg, real_t lambda, real_t alpha) { void MLPPGAN::add_output_layer(MLPPUtilities::WeightDistributionType weight_init, MLPPReg::RegularizationType reg, real_t lambda, real_t alpha) {
MLPPLinAlg alg; MLPPLinAlg alg;
if (!_network.empty()) { if (!_network.empty()) {
_output_layer = new MLPPOldOutputLayer(_network[_network.size() - 1].n_hidden, "Sigmoid", "LogLoss", _network[_network.size() - 1].a, weight_init, reg, lambda, alpha); _output_layer = Ref<MLPPOutputLayer>(memnew(MLPPOutputLayer(_network.write[_network.size() - 1]->get_n_hidden(), MLPPActivation::ACTIVATION_FUNCTION_SIGMOID, _network.write[_network.size() - 1]->get_a(), weight_init, reg, lambda, alpha)));
} else { } else {
_output_layer = new MLPPOldOutputLayer(_k, "Sigmoid", "LogLoss", alg.gaussianNoise(_n, _k), weight_init, reg, lambda, alpha); _output_layer = Ref<MLPPOutputLayer>(memnew(MLPPOutputLayer(_k, MLPPActivation::ACTIVATION_FUNCTION_SIGMOID, alg.gaussian_noise(_n, _k), weight_init, reg, lambda, alpha)));
} }
} }
MLPPGAN::MLPPGAN(real_t k, std::vector<std::vector<real_t>> output_set) { MLPPGAN::MLPPGAN(real_t k, const Ref<MLPPMatrix> &output_set) {
_output_set = output_set; _output_set = output_set;
_n = _output_set.size(); _n = _output_set->size().y;
_k = k; _k = k;
} }
@ -150,183 +158,210 @@ MLPPGAN::MLPPGAN() {
} }
MLPPGAN::~MLPPGAN() { MLPPGAN::~MLPPGAN() {
delete _output_layer;
} }
std::vector<std::vector<real_t>> MLPPGAN::model_set_test_generator(std::vector<std::vector<real_t>> X) { Ref<MLPPMatrix> MLPPGAN::model_set_test_generator(const Ref<MLPPMatrix> &X) {
if (!_network.empty()) { if (!_network.empty()) {
_network[0].input = X; _network.write[0]->set_input(X);
_network[0].forwardPass(); _network.write[0]->forward_pass();
for (uint32_t i = 1; i <= _network.size() / 2; i++) { for (int i = 1; i <= _network.size() / 2; i++) {
_network[i].input = _network[i - 1].a; _network.write[i]->set_input(_network.write[i - 1]->get_a());
_network[i].forwardPass(); _network.write[i]->forward_pass();
} }
} }
return _network[_network.size() / 2].a;
return _network.write[_network.size() / 2]->get_a();
} }
std::vector<real_t> MLPPGAN::model_set_test_discriminator(std::vector<std::vector<real_t>> X) { Ref<MLPPVector> MLPPGAN::model_set_test_discriminator(const Ref<MLPPMatrix> &X) {
if (!_network.empty()) { if (!_network.empty()) {
for (uint32_t i = _network.size() / 2 + 1; i < _network.size(); i++) { for (int i = _network.size() / 2 + 1; i < _network.size(); i++) {
if (i == _network.size() / 2 + 1) { if (i == _network.size() / 2 + 1) {
_network[i].input = X; _network.write[i]->set_input(X);
} else { } else {
_network[i].input = _network[i - 1].a; _network.write[i]->set_input(_network.write[i - 1]->get_a());
} }
_network[i].forwardPass(); _network.write[i]->forward_pass();
} }
_output_layer->input = _network[_network.size() - 1].a; _output_layer->set_input(_network.write[_network.size() - 1]->get_a());
} }
_output_layer->forwardPass(); _output_layer->forward_pass();
return _output_layer->a; return _output_layer->get_a();
} }
real_t MLPPGAN::cost(std::vector<real_t> y_hat, std::vector<real_t> y) { real_t MLPPGAN::cost(const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &y) {
MLPPReg regularization; MLPPReg regularization;
class MLPPCost cost; MLPPCost mlpp_cost;
real_t totalRegTerm = 0; real_t total_reg_term = 0;
auto cost_function = _output_layer->cost_map[_output_layer->cost];
if (!_network.empty()) { if (!_network.empty()) {
for (uint32_t i = 0; i < _network.size() - 1; i++) { for (int i = 0; i < _network.size() - 1; i++) {
totalRegTerm += regularization.regTerm(_network[i].weights, _network[i].lambda, _network[i].alpha, _network[i].reg); total_reg_term += regularization.reg_termm(_network.write[i]->get_weights(), _network.write[i]->get_lambda(), _network.write[i]->get_alpha(), _network.write[i]->get_reg());
} }
} }
return (cost.*cost_function)(y_hat, y) + totalRegTerm + regularization.regTerm(_output_layer->weights, _output_layer->lambda, _output_layer->alpha, _output_layer->reg); return mlpp_cost.run_cost_norm_vector(_output_layer->get_cost(), y_hat, y) + total_reg_term + regularization.reg_termv(_output_layer->get_weights(), _output_layer->get_lambda(), _output_layer->get_alpha(), _output_layer->get_reg());
} }
void MLPPGAN::forward_pass() { void MLPPGAN::forward_pass() {
MLPPLinAlg alg; MLPPLinAlg alg;
if (!_network.empty()) { if (!_network.empty()) {
_network[0].input = alg.gaussianNoise(_n, _k); _network.write[0]->set_input(alg.gaussian_noise(_n, _k));
_network[0].forwardPass(); _network.write[0]->forward_pass();
for (uint32_t i = 1; i < _network.size(); i++) { for (int i = 1; i < _network.size(); i++) {
_network[i].input = _network[i - 1].a; _network.write[i]->set_input(_network.write[i - 1]->get_a());
_network[i].forwardPass(); _network.write[i]->forward_pass();
} }
_output_layer->input = _network[_network.size() - 1].a; _output_layer->set_input(_network.write[_network.size() - 1]->get_a());
} else { // Should never happen, though. } else { // Should never happen, though.
_output_layer->input = alg.gaussianNoise(_n, _k); _output_layer->set_input(alg.gaussian_noise(_n, _k));
} }
_output_layer->forwardPass(); _output_layer->forward_pass();
_y_hat = _output_layer->a; _y_hat = _output_layer->get_a();
} }
void MLPPGAN::update_discriminator_parameters(std::vector<std::vector<std::vector<real_t>>> hidden_layer_updations, std::vector<real_t> output_layer_updation, real_t learning_rate) { void MLPPGAN::update_discriminator_parameters(const Vector<Ref<MLPPMatrix>> &hidden_layer_updations, const Ref<MLPPVector> &output_layer_updation, real_t learning_rate) {
MLPPLinAlg alg; MLPPLinAlg alg;
_output_layer->weights = alg.subtraction(_output_layer->weights, output_layer_updation); _output_layer->set_weights(alg.subtractionnv(_output_layer->get_weights(), output_layer_updation));
_output_layer->bias -= learning_rate * alg.sum_elements(_output_layer->delta) / _n; real_t output_layer_bias = _output_layer->get_bias();
output_layer_bias -= learning_rate * alg.sum_elementsv(_output_layer->get_delta()) / _n;
_output_layer->set_bias(output_layer_bias);
if (!_network.empty()) { if (!_network.empty()) {
_network[_network.size() - 1].weights = alg.subtraction(_network[_network.size() - 1].weights, hidden_layer_updations[0]); Ref<MLPPHiddenLayer> layer = _network[_network.size() - 1];
_network[_network.size() - 1].bias = alg.subtractMatrixRows(_network[_network.size() - 1].bias, alg.scalarMultiply(learning_rate / _n, _network[_network.size() - 1].delta));
for (int i = static_cast<int>(_network.size()) - 2; i > static_cast<int>(_network.size()) / 2; i--) { layer->set_weights(alg.subtractionm(layer->get_weights(), hidden_layer_updations[0]));
_network[i].weights = alg.subtraction(_network[i].weights, hidden_layer_updations[(_network.size() - 2) - i + 1]); layer->set_bias(alg.subtract_matrix_rows(layer->get_bias(), alg.scalar_multiplym(learning_rate / _n, layer->get_delta())));
_network[i].bias = alg.subtractMatrixRows(_network[i].bias, alg.scalarMultiply(learning_rate / _n, _network[i].delta));
for (int i = _network.size() - 2; i > _network.size() / 2; i--) {
layer = _network[i];
layer->set_weights(alg.subtractionm(layer->get_weights(), hidden_layer_updations[(_network.size() - 2) - i + 1]));
layer->set_bias(alg.subtract_matrix_rows(layer->get_bias(), alg.scalar_multiplym(learning_rate / _n, layer->get_delta())));
} }
} }
} }
void MLPPGAN::update_generator_parameters(std::vector<std::vector<std::vector<real_t>>> hidden_layer_updations, real_t learning_rate) { void MLPPGAN::update_generator_parameters(const Vector<Ref<MLPPMatrix>> &hidden_layer_updations, real_t learning_rate) {
MLPPLinAlg alg; MLPPLinAlg alg;
if (!_network.empty()) { if (!_network.empty()) {
for (int i = _network.size() / 2; i >= 0; i--) { for (int i = _network.size() / 2; i >= 0; i--) {
Ref<MLPPHiddenLayer> layer = _network[i];
//std::cout << network[i].weights.size() << "x" << network[i].weights[0].size() << std::endl; //std::cout << network[i].weights.size() << "x" << network[i].weights[0].size() << std::endl;
//std::cout << hidden_layer_updations[(network.size() - 2) - i + 1].size() << "x" << hidden_layer_updations[(network.size() - 2) - i + 1][0].size() << std::endl; //std::cout << hidden_layer_updations[(network.size() - 2) - i + 1].size() << "x" << hidden_layer_updations[(network.size() - 2) - i + 1][0].size() << std::endl;
_network[i].weights = alg.subtraction(_network[i].weights, hidden_layer_updations[(_network.size() - 2) - i + 1]); layer->set_weights(alg.subtractionm(layer->get_weights(), hidden_layer_updations[(_network.size() - 2) - i + 1]));
_network[i].bias = alg.subtractMatrixRows(_network[i].bias, alg.scalarMultiply(learning_rate / _n, _network[i].delta)); layer->set_bias(alg.subtract_matrix_rows(layer->get_bias(), alg.scalar_multiplym(learning_rate / _n, layer->get_delta())));
} }
} }
} }
std::tuple<std::vector<std::vector<std::vector<real_t>>>, std::vector<real_t>> MLPPGAN::compute_discriminator_gradients(std::vector<real_t> y_hat, std::vector<real_t> _output_set) { MLPPGAN::ComputeDiscriminatorGradientsResult MLPPGAN::compute_discriminator_gradients(const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &output_set) {
class MLPPCost cost; MLPPCost mlpp_cost;
MLPPActivation avn; MLPPActivation avn;
MLPPLinAlg alg; MLPPLinAlg alg;
MLPPReg regularization; MLPPReg regularization;
std::vector<std::vector<std::vector<real_t>>> cumulativeHiddenLayerWGrad; // Tensor containing ALL hidden grads. ComputeDiscriminatorGradientsResult res;
auto costDeriv = _output_layer->costDeriv_map[_output_layer->cost]; Ref<MLPPVector> cost_deriv = mlpp_cost.run_cost_deriv_vector(_output_layer->get_cost(), y_hat, _output_set);
auto outputAvn = _output_layer->activation_map[_output_layer->activation]; Ref<MLPPVector> activ_deriv = avn.run_activation_deriv_vector(_output_layer->get_activation(), _output_layer->get_z());
_output_layer->delta = alg.hadamard_product((cost.*costDeriv)(y_hat, _output_set), (avn.*outputAvn)(_output_layer->z, 1));
std::vector<real_t> outputWGrad = alg.mat_vec_mult(alg.transpose(_output_layer->input), _output_layer->delta); _output_layer->set_delta(alg.hadamard_productnv(cost_deriv, activ_deriv));
outputWGrad = alg.addition(outputWGrad, regularization.regDerivTerm(_output_layer->weights, _output_layer->lambda, _output_layer->alpha, _output_layer->reg));
res.output_w_grad = alg.mat_vec_multv(alg.transposem(_output_layer->get_input()), _output_layer->get_delta());
res.output_w_grad = alg.additionnv(res.output_w_grad, regularization.reg_deriv_termv(_output_layer->get_weights(), _output_layer->get_lambda(), _output_layer->get_alpha(), _output_layer->get_reg()));
if (!_network.empty()) { if (!_network.empty()) {
auto hiddenLayerAvn = _network[_network.size() - 1].activation_map[_network[_network.size() - 1].activation]; Ref<MLPPHiddenLayer> layer = _network[_network.size() - 1];
_network[_network.size() - 1].delta = alg.hadamard_product(alg.outerProduct(_output_layer->delta, _output_layer->weights), (avn.*hiddenLayerAvn)(_network[_network.size() - 1].z, 1)); Ref<MLPPVector> hidden_layer_activ_deriv = avn.run_activation_deriv_vector(layer->get_activation(), layer->get_z());
std::vector<std::vector<real_t>> hiddenLayerWGrad = alg.matmult(alg.transpose(_network[_network.size() - 1].input), _network[_network.size() - 1].delta);
cumulativeHiddenLayerWGrad.push_back(alg.addition(hiddenLayerWGrad, regularization.regDerivTerm(_network[_network.size() - 1].weights, _network[_network.size() - 1].lambda, _network[_network.size() - 1].alpha, _network[_network.size() - 1].reg))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well. layer->set_delta(alg.hadamard_productm(alg.outer_product(_output_layer->get_delta(), _output_layer->get_weights()), hidden_layer_activ_deriv));
Ref<MLPPMatrix> hidden_layer_w_grad = alg.matmultm(alg.transposem(layer->get_input()), layer->get_delta());
//std::cout << "HIDDENLAYER FIRST:" << hiddenLayerWGrad.size() << "x" << hiddenLayerWGrad[0].size() << std::endl; res.cumulative_hidden_layer_w_grad.push_back(alg.additionm(hidden_layer_w_grad, regularization.reg_deriv_termm(layer->get_weights(), layer->get_lambda(), layer->get_alpha(), layer->get_reg()))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well.
//std::cout << "WEIGHTS SECOND:" << network[network.size() - 1].weights.size() << "x" << network[network.size() - 1].weights[0].size() << std::endl;
for (int i = static_cast<int>(_network.size()) - 2; i > static_cast<int>(_network.size()) / 2; i--) { for (int i = static_cast<int>(_network.size()) - 2; i > static_cast<int>(_network.size()) / 2; i--) {
hiddenLayerAvn = _network[i].activation_map[_network[i].activation]; layer = _network[i];
_network[i].delta = alg.hadamard_product(alg.matmult(_network[i + 1].delta, alg.transpose(_network[i + 1].weights)), (avn.*hiddenLayerAvn)(_network[i].z, 1)); Ref<MLPPHiddenLayer> next_layer = _network[i + 1];
hiddenLayerWGrad = alg.matmult(alg.transpose(_network[i].input), _network[i].delta);
cumulativeHiddenLayerWGrad.push_back(alg.addition(hiddenLayerWGrad, regularization.regDerivTerm(_network[i].weights, _network[i].lambda, _network[i].alpha, _network[i].reg))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well. hidden_layer_activ_deriv = avn.run_activation_deriv_vector(layer->get_activation(), layer->get_z());
layer->set_delta(alg.hadamard_productm(alg.matmultm(next_layer->get_delta(), alg.transposem(next_layer->get_weights())), hidden_layer_activ_deriv));
hidden_layer_w_grad = alg.matmultm(alg.transposem(layer->get_input()), layer->get_delta());
res.cumulative_hidden_layer_w_grad.push_back(alg.additionm(hidden_layer_w_grad, regularization.reg_deriv_termm(layer->get_weights(), layer->get_lambda(), layer->get_alpha(), layer->get_reg()))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well.
} }
} }
return { cumulativeHiddenLayerWGrad, outputWGrad };
return res;
} }
std::vector<std::vector<std::vector<real_t>>> MLPPGAN::compute_generator_gradients(std::vector<real_t> y_hat, std::vector<real_t> _output_set) { Vector<Ref<MLPPMatrix>> MLPPGAN::compute_generator_gradients(const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &output_set) {
class MLPPCost cost; MLPPCost mlpp_cost;
MLPPActivation avn; MLPPActivation avn;
MLPPLinAlg alg; MLPPLinAlg alg;
MLPPReg regularization; MLPPReg regularization;
std::vector<std::vector<std::vector<real_t>>> cumulativeHiddenLayerWGrad; // Tensor containing ALL hidden grads. Vector<Ref<MLPPMatrix>> cumulative_hidden_layer_w_grad; // Tensor containing ALL hidden grads.
auto costDeriv = _output_layer->costDeriv_map[_output_layer->cost]; Ref<MLPPVector> cost_deriv = mlpp_cost.run_cost_deriv_vector(_output_layer->get_cost(), y_hat, _output_set);
auto outputAvn = _output_layer->activation_map[_output_layer->activation]; Ref<MLPPVector> activ_deriv = avn.run_activation_deriv_vector(_output_layer->get_activation(), _output_layer->get_z());
_output_layer->delta = alg.hadamard_product((cost.*costDeriv)(y_hat, _output_set), (avn.*outputAvn)(_output_layer->z, true));
std::vector<real_t> outputWGrad = alg.mat_vec_mult(alg.transpose(_output_layer->input), _output_layer->delta); _output_layer->set_delta(alg.hadamard_productnv(cost_deriv, activ_deriv));
outputWGrad = alg.addition(outputWGrad, regularization.regDerivTerm(_output_layer->weights, _output_layer->lambda, _output_layer->alpha, _output_layer->reg));
Ref<MLPPVector> output_w_grad = alg.mat_vec_multv(alg.transposem(_output_layer->get_input()), _output_layer->get_delta());
output_w_grad = alg.additionnv(output_w_grad, regularization.reg_deriv_termv(_output_layer->get_weights(), _output_layer->get_lambda(), _output_layer->get_alpha(), _output_layer->get_reg()));
if (!_network.empty()) { if (!_network.empty()) {
auto hiddenLayerAvn = _network[_network.size() - 1].activation_map[_network[_network.size() - 1].activation]; Ref<MLPPHiddenLayer> layer = _network[_network.size() - 1];
_network[_network.size() - 1].delta = alg.hadamard_product(alg.outerProduct(_output_layer->delta, _output_layer->weights), (avn.*hiddenLayerAvn)(_network[_network.size() - 1].z, 1));
std::vector<std::vector<real_t>> hiddenLayerWGrad = alg.matmult(alg.transpose(_network[_network.size() - 1].input), _network[_network.size() - 1].delta); Ref<MLPPVector> hidden_layer_activ_deriv = avn.run_activation_deriv_vector(layer->get_activation(), layer->get_z());
cumulativeHiddenLayerWGrad.push_back(alg.addition(hiddenLayerWGrad, regularization.regDerivTerm(_network[_network.size() - 1].weights, _network[_network.size() - 1].lambda, _network[_network.size() - 1].alpha, _network[_network.size() - 1].reg))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well.
layer->set_delta(alg.hadamard_productnv(alg.outer_product(_output_layer->get_delta(), _output_layer->get_weights()), hidden_layer_activ_deriv));
Ref<MLPPMatrix> hidden_layer_w_grad = alg.matmultm(alg.transposem(layer->get_input()), layer->get_delta());
cumulative_hidden_layer_w_grad.push_back(alg.additionm(hidden_layer_w_grad, regularization.reg_deriv_termm(layer->get_weights(), layer->get_lambda(), layer->get_alpha(), layer->get_reg()))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well.
for (int i = _network.size() - 2; i >= 0; i--) { for (int i = _network.size() - 2; i >= 0; i--) {
hiddenLayerAvn = _network[i].activation_map[_network[i].activation]; layer = _network[i];
_network[i].delta = alg.hadamard_product(alg.matmult(_network[i + 1].delta, alg.transpose(_network[i + 1].weights)), (avn.*hiddenLayerAvn)(_network[i].z, true)); Ref<MLPPHiddenLayer> next_layer = _network[i + 1];
hiddenLayerWGrad = alg.matmult(alg.transpose(_network[i].input), _network[i].delta);
cumulativeHiddenLayerWGrad.push_back(alg.addition(hiddenLayerWGrad, regularization.regDerivTerm(_network[i].weights, _network[i].lambda, _network[i].alpha, _network[i].reg))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well. hidden_layer_activ_deriv = avn.run_activation_deriv_vector(layer->get_activation(), layer->get_z());
layer->set_delta(alg.hadamard_productm(alg.matmultm(next_layer->get_delta(), alg.transposem(next_layer->get_weights())), hidden_layer_activ_deriv));
hidden_layer_w_grad = alg.matmultm(alg.transposem(layer->get_input()), layer->get_delta());
cumulative_hidden_layer_w_grad.push_back(alg.additionm(hidden_layer_w_grad, regularization.reg_deriv_termm(layer->get_weights(), layer->get_lambda(), layer->get_alpha(), layer->get_reg()))); // Adding to our cumulative hidden layer grads. Maintain reg terms as well.
} }
} }
return cumulativeHiddenLayerWGrad; return cumulative_hidden_layer_w_grad;
} }
void MLPPGAN::print_ui(int epoch, real_t cost_prev, std::vector<real_t> y_hat, std::vector<real_t> _output_set) { void MLPPGAN::print_ui(int epoch, real_t cost_prev, const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &output_set) {
MLPPUtilities::CostInfo(epoch, cost_prev, cost(y_hat, _output_set)); MLPPUtilities::cost_info(epoch, cost_prev, cost(y_hat, _output_set));
std::cout << "Layer " << _network.size() + 1 << ": " << std::endl;
MLPPUtilities::UI(_output_layer->weights, _output_layer->bias); PLOG_MSG("Layer " + itos(_network.size() + 1) + ": ");
MLPPUtilities::print_ui_vb(_output_layer->get_weights(), _output_layer->get_bias());
if (!_network.empty()) { if (!_network.empty()) {
for (int i = _network.size() - 1; i >= 0; i--) { for (int i = _network.size() - 1; i >= 0; i--) {
std::cout << "Layer " << i + 1 << ": " << std::endl; Ref<MLPPHiddenLayer> layer = _network[i];
MLPPUtilities::UI(_network[i].weights, _network[i].bias);
PLOG_MSG("Layer " + itos(i + 1) + ": ");
MLPPUtilities::print_ui_mb(layer->get_weights(), layer->get_bias());
} }
} }
} }

47
mlpp/gan/gan.h
View File

@ -15,12 +15,8 @@
#include "../hidden_layer/hidden_layer.h" #include "../hidden_layer/hidden_layer.h"
#include "../output_layer/output_layer.h" #include "../output_layer/output_layer.h"
#include "../hidden_layer/hidden_layer_old.h" #include "../activation/activation.h"
#include "../output_layer/output_layer_old.h" #include "../utilities/utilities.h"
#include <string>
#include <tuple>
#include <vector>
class MLPPGAN : public Reference { class MLPPGAN : public Reference {
GDCLASS(MLPPGAN, Reference); GDCLASS(MLPPGAN, Reference);
@ -37,45 +33,50 @@ public:
void set_k(const int val); void set_k(const int val);
*/ */
std::vector<std::vector<real_t>> generate_example(int n); Ref<MLPPMatrix> generate_example(int n);
void gradient_descent(real_t learning_rate, int max_epoch, bool ui = false); void gradient_descent(real_t learning_rate, int max_epoch, bool ui = false);
real_t score(); real_t score();
void save(std::string file_name); void save(const String &file_name);
void add_layer(int n_hidden, std::string activation, std::string weight_init = "Default", std::string reg = "None", real_t lambda = 0.5, real_t alpha = 0.5); void add_layer(int n_hidden, MLPPActivation::ActivationFunction activation, MLPPUtilities::WeightDistributionType weight_init = MLPPUtilities::WEIGHT_DISTRIBUTION_TYPE_DEFAULT, MLPPReg::RegularizationType reg = MLPPReg::REGULARIZATION_TYPE_NONE, real_t lambda = 0.5, real_t alpha = 0.5);
void add_output_layer(std::string weight_init = "Default", std::string reg = "None", real_t lambda = 0.5, real_t alpha = 0.5); void add_output_layer(MLPPUtilities::WeightDistributionType weight_init = MLPPUtilities::WEIGHT_DISTRIBUTION_TYPE_DEFAULT, MLPPReg::RegularizationType reg = MLPPReg::REGULARIZATION_TYPE_NONE, real_t lambda = 0.5, real_t alpha = 0.5);
MLPPGAN(real_t k, std::vector<std::vector<real_t>> output_set); MLPPGAN(real_t k, const Ref<MLPPMatrix> &output_set);
MLPPGAN(); MLPPGAN();
~MLPPGAN(); ~MLPPGAN();
protected: protected:
std::vector<std::vector<real_t>> model_set_test_generator(std::vector<std::vector<real_t>> X); // Evaluator for the generator of the gan. Ref<MLPPMatrix> model_set_test_generator(const Ref<MLPPMatrix> &X); // Evaluator for the generator of the gan.
std::vector<real_t> model_set_test_discriminator(std::vector<std::vector<real_t>> X); // Evaluator for the discriminator of the gan. Ref<MLPPVector> model_set_test_discriminator(const Ref<MLPPMatrix> &X); // Evaluator for the discriminator of the gan.
real_t cost(std::vector<real_t> y_hat, std::vector<real_t> y); real_t cost(const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &y);
void forward_pass(); void forward_pass();
void update_discriminator_parameters(std::vector<std::vector<std::vector<real_t>>> hidden_layer_updations, std::vector<real_t> output_layer_updation, real_t learning_rate); void update_discriminator_parameters(const Vector<Ref<MLPPMatrix>> &hidden_layer_updations, const Ref<MLPPVector> &output_layer_updation, real_t learning_rate);
void update_generator_parameters(std::vector<std::vector<std::vector<real_t>>> hidden_layer_updations, real_t learning_rate); void update_generator_parameters(const Vector<Ref<MLPPMatrix>> &hidden_layer_updations, real_t learning_rate);
std::tuple<std::vector<std::vector<std::vector<real_t>>>, std::vector<real_t>> compute_discriminator_gradients(std::vector<real_t> y_hat, std::vector<real_t> output_set); struct ComputeDiscriminatorGradientsResult {
std::vector<std::vector<std::vector<real_t>>> compute_generator_gradients(std::vector<real_t> y_hat, std::vector<real_t> output_set); Vector<Ref<MLPPMatrix>> cumulative_hidden_layer_w_grad; // Tensor containing ALL hidden grads.
Ref<MLPPVector> output_w_grad;
};
void print_ui(int epoch, real_t cost_prev, std::vector<real_t> y_hat, std::vector<real_t> output_set); ComputeDiscriminatorGradientsResult compute_discriminator_gradients(const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &output_set);
Vector<Ref<MLPPMatrix>> compute_generator_gradients(const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &output_set);
void print_ui(int epoch, real_t cost_prev, const Ref<MLPPVector> &y_hat, const Ref<MLPPVector> &output_set);
static void _bind_methods(); static void _bind_methods();
std::vector<std::vector<real_t>> _output_set; Ref<MLPPMatrix> _output_set;
std::vector<real_t> _y_hat; Ref<MLPPVector> _y_hat;
std::vector<MLPPOldHiddenLayer> _network; Vector<Ref<MLPPHiddenLayer>> _network;
MLPPOldOutputLayer *_output_layer; Ref<MLPPOutputLayer> _output_layer;
int _n; int _n;
int _k; int _k;