encoder.hpp

Go to the documentation of this file.

/*************************************************************************
 *
 *  Project
 *                         _____ _____  __  __ _____
 *                        / ____|  __ \|  \/  |  __ \
 *  ___  _ __   ___ _ __ | |  __| |__) | \  / | |__) |
 * / _ \| '_ \ / _ \ '_ \| | |_ |  ___/| |\/| |  ___/
 *| (_) | |_) |  __/ | | | |__| | |    | |  | | |
 * \___/| .__/ \___|_| |_|\_____|_|    |_|  |_|_|
 *      | |
 *      |_|
 *
 * Copyright (C) Akiel Aries, <akiel@akiel.org>, et al.
 *
 * This software is licensed as described in the file LICENSE, which
 * you should have received as part of this distribution. The terms
 * among other details are referenced in the official documentation
 * seen here : https://akielaries.github.io/openGPMP/ along with
 * important files seen in this project.
 *
 * You may opt to use, copy, modify, merge, publish, distribute
 * and/or sell copies of the Software, and permit persons to whom
 * the Software is furnished to do so, under the terms of the
 * LICENSE file. As this is an Open Source effort, all implementations
 * must be of the same methodology.
 *
 *
 *
 * This software is distributed on an AS IS basis, WITHOUT
 * WARRANTY OF ANY KIND, either express or implied.
 *
 ************************************************************************/
 
#ifndef ENCODER_HPP
#define ENCODER_HPP
 
#include <string>
#include <tuple>
#include <vector>
 
namespace gpmp {
 
namespace ml {
 
// TODO: this base class could probably use its own gradient descent method to
// use across all inheriting classes OR create a class of trainers to use in
// `trainers.cpp`?
class AutoEncoder {
  public:
    int input_size;
 
    int hidden_size;
 
    int output_size;
 
    double learning_rate;
 
    std::vector<std::vector<double>> weights_input_hidden;
 
    std::vector<std::vector<double>> weights_hidden_output;
 
    AutoEncoder(int input_size,
                int hidden_size,
                int output_size,
                double learning_rate);
 
    std::vector<double> sigmoid(const std::vector<double> &x);
 
    std::vector<double> forward(const std::vector<double> &input);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs);
 
    void lrate_set(double initial_rate);
 
    virtual void lrate_update(int epoch);
 
    void display();
 
    virtual void save(const std::string &filename) const;
 
    virtual void load(const std::string &filename);
};
 
class SparseAutoEncoder : public AutoEncoder {
  public:
    double sparsity_weight;
 
    double sparsity_target;
 
    SparseAutoEncoder(int input_size,
                      int hidden_size,
                      int output_size,
                      double learning_rate,
                      double sparsity_weight,
                      double sparsity_target);
 
    void train(const std::vector<std::vector<double>> &training_data,
               int epochs) override;
};
 
class DenoisingAutoEncoder : public AutoEncoder {
  public:
    double corruption_level;
 
    DenoisingAutoEncoder(int input_size,
                         int hidden_size,
                         int output_size,
                         double learning_rate,
                         double corruption_level);
 
    void train(const std::vector<std::vector<double>> &training_data,
               int epochs) override;
};
 
class ContractiveAutoEncoder : public AutoEncoder {
  public:
    double contractive_weight;
 
    ContractiveAutoEncoder(int input_size,
                           int hidden_size,
                           int output_size,
                           double learning_rate,
                           double contractive_weight);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs) override;
};
 
class MDLAutoEncoder : public AutoEncoder {
  public:
    double mdl_weight;
 
    MDLAutoEncoder(int input_size,
                   int hidden_size,
                   int output_size,
                   double learning_rate,
                   double mdl_weight);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs) override;
};
 
class ConcreteAutoEncoder : public AutoEncoder {
  public:
    double temperature;
 
    ConcreteAutoEncoder(int input_size,
                        int hidden_size,
                        int output_size,
                        double learning_rate,
                        double temperature);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs) override;
};
 
class VariationalAutoEncoder : public AutoEncoder {
  public:
    std::vector<double> hidden_mean;
    std::vector<double> hidden_log_variance;
 
    double sample_dist();
 
    double reparameterize(double mean, double log_variance);
 
    std::vector<double> encoder(const std::vector<double> &input);
 
    std::vector<double> decoder(const std::vector<double> &hidden_sampled);
 
    void gradient_descent(const std::vector<double> &input,
                          const std::vector<double> &output,
                          const std::vector<double> &hidden_sampled);
 
    VariationalAutoEncoder(int input_size,
                           int hidden_size,
                           int output_size,
                           double learning_rate);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs) override;
};
 
class RecurrentAutoEncoder : public AutoEncoder {
  public:
    std::vector<std::vector<double>> weights_recurrent;
 
    RecurrentAutoEncoder(int input_size,
                         int hidden_size,
                         int output_size,
                         double learning_rate);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs) override;
 
    std::vector<double> recurr_fwd(const std::vector<double> &input,
                                   const std::vector<double> &previous_hidden);
};
 
class FullAutoEncoder : public AutoEncoder {
  public:
    FullAutoEncoder(int input_size,
                    int hidden_size,
                    int output_size,
                    double learning_rate);
 
    virtual void train(const std::vector<std::vector<double>> &training_data,
                       int epochs) override;
};
 
} // namespace ml
 
} // namespace gpmp
 
#endif