knn.rar_SOM_knn

/* * Copyright (c) 2006, Seweryn Habdank-Wojewodzki * Copyright (c) 2006, Janusz Rybarski * * All rights reserved. * * Redistribution and use in source and binary forms, * with or without modification, are permitted provided * that the following conditions are met: * * Redistributions of source code must retain the above * copyright notice, this list of conditions and the * following disclaimer. * * 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. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS * AND CONTRIBUTORS "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 COPYRIGHT OWNER OR CONTRIBUTORS 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. */ /* * e-mail: habdank AT gmail DOT com * e-mail: janusz.rybarski AT gmail DOT com * * File created: Thu 10 Apr 2006 11:05:06 CEST * Last modified: Wed 08 Aug 2007 18:22:31 CEST */ #include <boost/cstdint.hpp> #include "../include/debugger.hpp" #include "../include/std_headers.hpp" #include "../include/neural_net_headers.hpp" #include "../include/data_parser.hpp" #include "configuration.hpp" // PMAP is used just to hold program for debugging memory usage #ifdef PMAP #include <cstdio> #endif //PMAP /** * \mainpage Kohonen Neural Network Library Demo * * \par Program is designed for making pattern recognition using Kohonen Self-oragnizing Map. * * \section intro_sec Introduction * * \section options Options * * \par Program usage: * * knnl_demo [OPTIONS] [FILE] * * If FILE is not given then standard input (stdin) is taken as input stream. * * \par Generic options: * * - -h help message * * - -V version string * * - -r number of rows in neural matrix * * - -c number of columns in neural martix * * - -e number of epochs that training process will pass * * \par Hidden options: * * - -i [ --input-file ] arg input file name * * \author Seweryn Habdank-Wojewodzki * \author Janusz Rybarski */ /** * \file main.cpp * \brief Main file of the program contains ::boost::int32_t main ( ::boost::int32_t argc, char * argv[] ) function :-) */ /** * \var DEBUGGER_STREAM * \brief Set dynamical pointer to the stream. * \code * DEBUGGER_STREAM = * ::std::auto_ptr < ::std::ofstream > ( new ::std::ofstream ( "_debugger.out" ) ); * \endcode */ #ifdef FTDEBUG ::std::auto_ptr < ::std::ofstream > DEBUGGER_STREAM; #endif //FTDEGUG /** * \defgroup data_storages Data storages */ /** * Configuration. * \ingroup data_storages */ ::config::Configuration CONF; /** * Main function of the program. */ ::boost::int32_t main ( ::boost::int32_t argc, char * argv[] ) { #ifdef FTDEBUG DEBUGGER_STREAM = ::std::auto_ptr < ::std::ofstream > ( new ::std::ofstream ( "_debugger.out" ) ); #endif //FTDEBUG // set program name and version number CONF.set_program_name ( "Kohonen Network Library Demo" ); CONF.set_version_number ( "0.0.1" ); ::boost::int32_t return_value = 1; // flag used when compilation is prepared for the tests using pmap linux command #ifdef PMAP ::std::cout << "Press any key."; getchar (); #endif //PMAP try { // try to read command line return_value = ::config::do_program_options ( argc, argv, CONF ); // if PROCEED was returned then follow the white rabbit if ( return_value != ::config::PROCEED ) { return return_value; } // set NULL to the istream pointer ::std::istream * file_ptr = static_cast < ::std::istream * > ( 0 ); // if input_file_name is empty set standard input as input for the program // this is important when program have to work as filter for the stream console if ( CONF.parameters.input_file_name == "" ) { file_ptr = new ::std::istream ( ::std::cin.rdbuf() ) ; // ::std::cout << "No file name specified." << ::std::endl; } else { // set real file as input stream file_ptr = new ::std::ifstream ( CONF.parameters.input_file_name.c_str() ); // ::std::cout << CONF.parameters.input_file_name << ::std::endl; if ( !file_ptr ) { ::std::cout << "Error in reading file." << ::std::endl; ::std::exit ( EXIT_FAILURE ); } } // read file for very big files if could cause problems, // but for files smaller than available memory this is faster solution ::std::stringstream tmp_stream; tmp_stream << file_ptr->rdbuf(); delete file_ptr; // start random generator ::std::srand (static_cast<unsigned int> (time (NULL))); // it could cause problems try { // set number of rows for network ::boost::int32_t R = CONF.parameters.no_rows; // set number of columns for network ::boost::int32_t C = CONF.parameters.no_columns; // set number of the the training epochs // one epoch it is training through all data stored in container ::boost::int32_t const no_epochs = CONF.parameters.no_epochs; // example of using macro for debugging D ( R ); D ( C ); D ( no_epochs ); typedef double data_type; // type of the single data vector // for example if we have pairs of the data like: // { (1,2.4), (2,3.2), (3,4.7) } // data_type describes a type of the single element // V_d describes type of (1,2.4) typedef ::std::vector < data_type > V_d; // type below describes type of the container of all data typedef ::std::vector < V_d > V_v_d; // list container is quite good for storing data, // but thare is no possibility to use ::std::random_shuffle() //typedef ::std::list < V_d > V_v_d; V_v_d data; ::data_parser::Data_parser < V_v_d > data_parser; data_parser ( tmp_stream, data ); D ( data.size() ); // configure Cauchy hat function typedef ::neural_net::Cauchy_function < V_d::value_type, V_d::value_type, ::boost::int32_t > C_a_f; // Activation function for neurons C_a_f c_a_f ( 2.0, 1 ); // configure Gauss hat function typedef ::neural_net::Gauss_function < V_d::value_type, V_d::value_type, ::boost::int32_t > G_a_f; // Activation function for neurons G_a_f g_a_f ( 2.0, 1 ); // prepare Euclidean distance function typedef ::distance::Euclidean_distance_function < V_d > E_d_t; E_d_t e_d; // and weighted Euclidean distance function typedef ::distance::Weighted_euclidean_distance_function < V_d, V_d > We_d_t; V_d coefs; ::neural_net::Ranges < V_v_d > data_ranges ( *data.begin() ); data_ranges ( data ); //preparing proper parameters for weighted distance V_d::const_iterator pos_max = data_ranges.get_max().begin(); V_d::const_iterator pos_min = data_ranges.get_min().begin(); for ( ;