view toolboxes/RBM/Main.m @ 0:e9a9cd732c1e tip

first hg version after svn

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%  
% Main script of KRBM-Knowledge-based Restricted Boltzmann Machine   %
% -*-sontran2012-*-                                                  %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% loading background knowledge 
% bk_file  = 'XOR.bk';
% w = 5;
W  = rand(0,0);
mW = rand(0,0);
% [W mW nNames] = encode_bk(bk_file,w)
%%% visualize_system(W,nNames,w,100,100);
%% Loading data
PRO_DIR = 'C://Pros/';
% dat_file = 'XOR.dat';
% dat_file = 'DNA.dat';
% train_file  = strcat(PRO_DIR,'/Data/ICDAR/digit_train_data.mat';
% train_label = strcat(PRO_DIR,'/Data/ICDAR/digit_train_label.mat';
% test_file   = strcat(PRO_DIR,'/Data/ICDAR/digit_test_data.mat';
% test_label  = strcat(PRO_DIR,'/Data/ICDAR/digit_test_label.mat';
%-----------------------------------------------------
% train_file  = strcat(PRO_DIR,'/Data/MNIST/mnist_train_data.mat';
% train_label = strcat(PRO_DIR,'/Data/MNIST/mnist_train_label.mat';
% test_file   = strcat(PRO_DIR,'/Data/MNIST/mnist_test_data.mat';
% test_label  = strcat(PRO_DIR,'/Data/MNIST/mnist_test_label.mat';
%-----------------------------------------------------
% train_file  = strcat(PRO_DIR,'/RBMLIB/train_dat.mat');
% train_label = strcat(PRO_DIR,'/RBMLIB/train_lab.mat');
% test_file   = strcat(PRO_DIR,'/RBMLIB/test_dat.mat');
% test_label  = strcat(PRO_DIR,'/RBMLIB/test_lab.mat');
%-----------------------------------------------------
% train_file  = strcat(PRO_DIR,'/RBMLIB/train_dat.mat');
data_file = strcat(PRO_DIR,'/DATA/XOR/xorall.mat');
mod_file = strcat(PRO_DIR,'/DATA/XOR/model.mat');
%% Move weights encoded from rule to right position of input if needed

%% Setting parameters
conf.hidNum = 4;
conf.eNum   = 50;                                                           % number of epoch
conf.bNum   = 1;                                                           % number of batches devides the data set
conf.sNum   = 4;                                                          % number of samples per batch
conf.gNum   = 1;                                                            % number of iteration for gibb sampling
conf.params = [0.5 0.5 0.1 0.00002];                                        % [lr1 lr2 momentum cost] 
                                                                            % lr1: learning rate in first 10 epoches
                                                                            % lr2: learning rate in the other epoches
%% Training system
% training_kbrm(conf,W,mW,dat_file);                                        % unsupevised learning
% [W visB hidB] = gen_training_kbrbm(conf,W,mW,train_file,train_label);     % unsupervised learning for generative classification
% dis_training_kbrm(conf,W,mW,dat_file,dat_label);                          % unsupervised learning for discriminative classification
[W visB hidB] = training_rbm(conf,W,mW,data_file);
save(mod_file,'W','mW','visB','hidB');
nNames = {'X' 'Y' 'Z'};
visualize_system(W,nNames,100,100);
%% Classify (if wanted)
% gen_classify(W, visB, hidB,mW,test_file,test_label);
%% Generate features (if wanted)
% in_f_file = '../Data/ICDAR/images.bin';
% out_f_file = '../Data/ICDAR/feature.mat';
% input_features = mnist2mat(in_f_file);
% output_features = input_features*W;
% save(out_f_file,'output_features');
%% Extract knowledge (if wanted)
% name_file = 'names.txt';                                                    % File contains name of input
% rule_file = 'rules.txt';
% extract_knowledge(rule_file,W);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% END%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear;