深度学习DBN代码

深度学习，DBN模型代码，c语言，代码完整，由matlab转写过来

#include 
#include 
#include 
#include “Hiddenlayer.h“
#include “RBM.h“
#include “LogisticRegression.h“
#include “DBN.h“
#include “utils.h“

void test_dbn（void）;


double uniform（double min double max） {
  return rand（） / （RAND_MAX + 1.0） * （max - min） + min;  
}

int binomial（int n double p） {
  if（p < 0 || p > 1） return 0;

  int i;
  int c = 0;
  double r;

  for（i=0; i    r = rand（） / （RAND_MAX + 1.0）;
    if （r < p） c++;
  }

  return c;
}

double sigmoid（double x） {
  return 1.0 / （1.0 + exp（-x））;
}



// DBN
void DBN__construct（DBN* this int N \
                    int n_ins int *hidden_layer_sizes int n_outs int n_layers） {
  int i input_size;

  this->N = N;
  this->n_ins = n_ins;
  this->hidden_layer_sizes = hidden_layer_sizes;
  this->n_outs = n_outs;
  this->n_layers = n_layers;

  this->sigmoid_layers = （Hiddenlayer *）malloc（sizeof（Hiddenlayer） * n_layers）;
  this->rbm_layers = （RBM *）malloc（sizeof（RBM） * n_layers）;

  // construct multi-layer
  for（i=0; iyers; i++） {
    if（i == 0） {
      input_size = n_ins;
    } else {
      input_size = hidden_layer_sizes[i-1];
    }

    // construct sigmoid_layer
    Hiddenlayer__construct（&（this->sigmoid_layers[i]） \
                           N input_size hidden_layer_sizes[i] NULL NULL）;

    // construct rbm_layer
    RBM__construct（&（this->rbm_layers[i]） N input_size hidden_layer_sizes[i] \
                   this->sigmoid_layers[i].W this->sigmoid_layers[i].b NULL）;
    
  }

  // layer for output using LogisticRegression
  LogisticRegression__construct（&（this->log_layer） \
                                N hidden_layer_sizes[n_layers-1] n_outs）;
  
}

void DBN__destruct（DBN* this） {
  int i;
  for（i=0; in_layers; i++） {
    Hiddenlayer__destruct（&（this->sigmoid_layers[i]））;
    RBM__destruct（&（this->rbm_layers[i]））;
  }
  free（this->sigmoid_layers）;
  free（this->rbm_layers）;
}

void DBN_pretrain（DBN* this int *input double lr int k int epochs） {
  int i j l m n epoch;
  
  int *layer_input;
  int prev_layer_input_size;
  int *prev_layer_input;

  int *train_X = （int *）malloc（sizeof（int） * this->n_ins）;

  for（i=0; in_layers; i++） { // layer-wise

    for（epoch=0; epoch
      for（n=0; nN; n++） { // input x1...xN
        // initial input
        for（m=0; mn_ins; m++） train_X[m] = input[n * this->n_ins + m];

        // layer input
        for（l=0; l<=i; l++） {
          if（l == 0） {
            layer_input = （int *）malloc（sizeof（int） * this->n_ins）;
            for（j=0; jn_ins; j++） layer_input[j] = train_X[j];
          } else {
            if（l == 1） prev_layer_input_size = this->n_ins;
            else prev_layer_input_size = this->hidden_layer_sizes[l-2];

            prev_layer_input = （int *）malloc（sizeof（int） * prev_layer_input_size）;
            for（j=0; jyer_input_size; j++） prev_layer_input[j] = layer_input[j];
            fr

 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----

     文件      48205  2013-04-30 19:36  Deep_Belief_Nets\bin\Debug\Deep_Belief_Nets.exe

     文件      15138  2013-03-27 08:13  Deep_Belief_Nets\DBN.c

     文件        440  2013-03-27 08:13  Deep_Belief_Nets\DBN.h

     文件       1273  2013-04-30 19:36  Deep_Belief_Nets\Deep_Belief_Nets.cbp

     文件        559  2013-04-30 20:35  Deep_Belief_Nets\Deep_Belief_Nets.depend

     文件        322  2013-04-30 20:36  Deep_Belief_Nets\Deep_Belief_Nets.layout

     文件        391  2013-03-27 08:13  Deep_Belief_Nets\Hiddenlayer.h

     文件        500  2013-03-27 08:13  Deep_Belief_Nets\LogisticRegression.h

     文件      23535  2013-04-30 19:36  Deep_Belief_Nets\obj\Debug\DBN.o

     文件        613  2013-03-27 08:13  Deep_Belief_Nets\RBM.h

     文件        124  2013-03-27 08:13  Deep_Belief_Nets\utils.h

     目录          0  2013-05-07 15:03  Deep_Belief_Nets\bin\Debug

     目录          0  2013-05-07 15:03  Deep_Belief_Nets\obj\Debug

     目录          0  2013-05-07 15:03  Deep_Belief_Nets\bin

     目录          0  2013-05-07 15:03  Deep_Belief_Nets\obj

     目录          0  2013-05-07 15:03  Deep_Belief_Nets

----------- ---------  ---------- -----  ----

                91100                    16