GaussianContinuum.h

Go to the documentation of this file.

// -*- C++ -*-

// GaussianContinuum.h
//
// Copyright (C) 2004 Yoshua Bengio & Martin Monperrus
// 
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
// 
//  1. Redistributions of source code must retain the above copyright
//     notice, this list of conditions and the following disclaimer.
// 
//  2. Redistributions in binary form must reproduce the above copyright
//     notice, this list of conditions and the following disclaimer in the
//     documentation and/or other materials provided with the distribution.
// 
//  3. The name of the authors may not be used to endorse or promote
//     products derived from this software without specific prior written
//     permission.
// 
// THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
// IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
// OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN
// NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
// TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
// 
// This file is part of the PLearn library. For more information on the PLearn
// library, go to the PLearn Web site at www.plearn.org

/* *******************************************************      
   * $Id: GaussianContinuum.h,v 1.4 2004/08/17 15:28:21 larocheh Exp $
   ******************************************************* */

// Authors: Yoshua Bengio & Martin Monperrus

#ifndef GaussianContinuum_INC
#define GaussianContinuum_INC
#include <plearn/io/PStream.h>
#include <plearn_learners/generic/PLearner.h>
#include <plearn/var/Func.h>
#include <plearn/opt/Optimizer.h>
#include <plearn_learners/distributions/PDistribution.h>

namespace PLearn {
using namespace std;

class GaussianContinuum: public PLearner
{

private:

  typedef PLearner inherited;
  
protected:
  // NON-OPTION FIELDS
  int n;
  Func cost_of_one_example;
  Func verify_gradient_func;
  Var x, noise_var; // input vector
  Var b, W, c, V, muV, smV, smb, snV, snb; // explicit view of the parameters (also in parameters field).
  //Var W_src, c_src, V_src, muV_src, smV_src, smb_src, snV_src, snb_src; 
  //VarArray mu_neighbors, sm_neighbors, sn_neighbors, hidden_neighbors, input_neighbors, index_neighbors, tangent_plane_neighbors;
  Var tangent_targets; // target for the tangent vectors for one example 
  Var tangent_plane;
  Var mu, sm, sn, mu_noisy; // parameters of the conditional models
  Var p_x, p_target, p_neighbors, target_index, neigbor_indexes;
  Var sum_nll;

  PP<PDistribution> dist;

  // Random walk fields
  Array<VMat> ith_step_generated_set;

  // p(x) computation fields
  VMat train_and_generated_set;
  TMat<int> train_nearest_neighbors;
  TMat<int> validation_nearest_neighbors;
  TVec< Mat > Bs, Fs;
  Mat mus;
  Vec sms;
  Vec sns;

  Mat Ut_svd, V_svd;  // for SVD computation
  Vec S_svd;      // idem
  Vec z, zm, zn, x_minus_neighbor, w;
  Vec t_row, neighbor_row;

  real best_validation_cost;

  // *********************
  // * protected options *
  // *********************

  // ### declare protected option fields (such as learnt parameters) here
  VarArray parameters;

public:

  // ************************
  // * public build options *
  // ************************

  // ### declare public option fields (such as build options) here

  bool use_noise;
  real noise;
  string noise_type;
  int n_random_walk_step;
  int n_random_walk_per_point;
  bool save_image_mat;
  VMat image_points_vmat;
  Mat image_points_mat;
  Mat image_prob_mat;
  TMat<int> image_nearest_neighbors;
  real upper_y;
  real lower_y;
  real upper_x;
  real lower_x;
  int points_per_dim;
  real min_sigma;
  real min_diff;
  bool print_parameters;
  bool sm_bigger_than_sn;
  bool use_best_model;
  int n_neighbors; // number of neighbors used for gradient descent
  int n_neighbors_density; // number of neighbors for the p(x) density estimation
  int n_dim; // number of reduced dimensions (number of tangent vectors to compute)
  int compute_cost_every_n_epochs;
  string variances_transfer_function; // "square", "exp" or "softplus"
  PP<Optimizer> optimizer; // to estimate the function that predicts local tangent vectors given the input
  Var embedding;
  Func output_f;
  Func output_f_all;
  Func predictor; // predicts everything about the gaussian
  Func projection_error_f; // map output to projection error
  Func noisy_data;

  // manual construction of the tangent_predictor
  string architecture_type; // "neural_network" or "linear" or "" or "embedding_neural_nework" or "embedding_quadratic" 
  string output_type; // "tangent_plane", "embedding", or "tangent_plane+embedding".
  int n_hidden_units;

  int batch_size;

  real norm_penalization; // penalizes sum_i (||f_i||^2-1)^2
  real svd_threshold;
  real projection_error_regularization; // term to add on linear system diagonal, to solve for subspace fitting

  // ****************
  // * Constructors *
  // ****************

  // (Make sure the implementation in the .cc
  // initializes all fields to reasonable default values)
  GaussianContinuum();


  // ********************
  // * PLearner methods *
  // ********************

private: 

  // (Please implement in .cc)
  void build_();

  void compute_train_and_validation_costs();

  void make_random_walk();

  void get_image_matrix(Mat& image, VMat image_points_vmat, int begin, string file_path);

protected: 
  
  // (Please implement in .cc)
  static void declareOptions(OptionList& ol);

public:

  // ************************
  // **** Object methods ****
  // ************************

  virtual void build();

  virtual void makeDeepCopyFromShallowCopy(map<const void*, void*>& copies);

  // Declares other standard object methods.
  // If your class is not instantiatable (it has pure virtual methods)
  // you should replace this by PLEARN_DECLARE_ABSTRACT_OBJECT_METHODS.
  PLEARN_DECLARE_OBJECT(GaussianContinuum);


  // **************************
  // **** PLearner methods ****
  // **************************

  virtual int outputsize() const;

  virtual void forget();
  virtual void initializeParams();

    
  virtual void train();


  // (PLEASE IMPLEMENT IN .cc)
  virtual void computeOutput(const Vec& input, Vec& output) const;

  // (PLEASE IMPLEMENT IN .cc)
  virtual void computeCostsFromOutputs(const Vec& input, const Vec& output, 
                                       const Vec& target, Vec& costs) const;
                                

  // (PLEASE IMPLEMENT IN .cc)
  virtual TVec<string> getTestCostNames() const;

  // (PLEASE IMPLEMENT IN .cc)
  virtual TVec<string> getTrainCostNames() const;


  // *** SUBCLASS WRITING: ***
  // While in general not necessary, in case of particular needs 
  // (efficiency concerns for ex) you may also want to overload
  // some of the following methods:
  // virtual void computeOutputAndCosts(const Vec& input, const Vec& target, Vec& output, Vec& costs) const;
  // virtual void computeCostsOnly(const Vec& input, const Vec& target, Vec& costs) const;
  // virtual void test(VMat testset, PP<VecStatsCollector> test_stats, VMat testoutputs=0, VMat testcosts=0) const;
  // virtual int nTestCosts() const;
  // virtual int nTrainCosts() const;

};

// Declares a few other classes and functions related to this class.
DECLARE_OBJECT_PTR(GaussianContinuum);
  
} // end of namespace PLearn

#endif

---