GMM.h

Go to the documentation of this file.
00001 /*
00002  * This program is free software; you can redistribute it and/or modify
00003  * it under the terms of the GNU General Public License as published by
00004  * the Free Software Foundation; either version 3 of the License, or
00005  * (at your option) any later version.
00006  *
00007  * Written (W) 2011 Alesis Novik
00008  * Copyright (C) 2011 Berlin Institute of Technology and Max-Planck-Society
00009  */
00010 #ifndef _GMM_H__
00011 #define _GMM_H__
00012 
00013 #include <shogun/lib/config.h>
00014 
00015 #ifdef HAVE_LAPACK
00016 
00017 #include <shogun/distributions/Distribution.h>
00018 #include <shogun/distributions/Gaussian.h>
00019 #include <shogun/lib/common.h>
00020 
00021 #include <vector>
00022 
00023 using namespace std;
00024 
00025 namespace shogun
00026 {
00040 class CGMM : public CDistribution
00041 {
00042     public:
00044         CGMM();
00050         CGMM(int32_t n, ECovType cov_type=FULL);
00057         CGMM(vector<CGaussian*> components, SGVector<float64_t> coefficients,
00058                 bool copy=false);
00059         virtual ~CGMM();
00060 
00062         void cleanup();
00063 
00070         virtual bool train(CFeatures* data=NULL);
00071 
00080         float64_t train_em(float64_t min_cov=1e-9, int32_t max_iter=1000,
00081                 float64_t min_change=1e-9);
00082 
00093         float64_t train_smem(int32_t max_iter=100, int32_t max_cand=5,
00094                 float64_t min_cov=1e-9, int32_t max_em_iter=1000,
00095                 float64_t min_change=1e-9);
00096 
00102         void max_likelihood(SGMatrix<float64_t> alpha, float64_t min_cov);
00103 
00108         virtual int32_t get_num_model_parameters();
00109 
00115         virtual float64_t get_log_model_parameter(int32_t num_param);
00116 
00123         virtual float64_t get_log_derivative(
00124             int32_t num_param, int32_t num_example);
00125 
00133         virtual float64_t get_log_likelihood_example(int32_t num_example);
00134 
00142         virtual float64_t get_likelihood_example(int32_t num_example);
00143 
00150         virtual SGVector<float64_t> get_nth_mean(int32_t num);
00151 
00157         virtual void set_nth_mean(SGVector<float64_t> mean, int32_t num);
00158 
00165         virtual SGMatrix<float64_t> get_nth_cov(int32_t num);
00166 
00172         virtual void set_nth_cov(SGMatrix<float64_t> cov, int32_t num);
00173 
00178         virtual SGVector<float64_t> get_coef();
00179 
00184         virtual void set_coef(const SGVector<float64_t> coefficients);
00185 
00190         virtual vector<CGaussian*> get_comp();
00191 
00196         virtual void set_comp(vector<CGaussian*> components);
00197 
00202         SGVector<float64_t> sample();
00203 
00209         SGVector<float64_t> cluster(SGVector<float64_t> point);
00210 
00212         virtual const char* get_name() const { return "GMM"; }
00213 
00214     private:
00221         SGMatrix<float64_t> alpha_init(SGMatrix<float64_t> init_means);
00222 
00224         void register_params();
00225 
00235         void partial_em(int32_t comp1, int32_t comp2, int32_t comp3,
00236                 float64_t min_cov, int32_t max_em_iter, float64_t min_change);
00237 
00238     protected:
00240         vector<CGaussian*> m_components;
00242         SGVector<float64_t> m_coefficients;
00243 };
00244 }
00245 #endif //HAVE_LAPACK
00246 #endif //_GMM_H__
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Defines

SHOGUN Machine Learning Toolbox - Documentation