SparseLinearKernel.cpp

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) 1999-2008 Soeren Sonnenburg
00008  * Copyright (C) 1999-2008 Fraunhofer Institute FIRST and Max-Planck-Society
00009  */
00010 
00011 #include "lib/common.h"
00012 #include "lib/io.h"
00013 #include "features/Features.h"
00014 #include "features/SparseFeatures.h"
00015 #include "kernel/SparseLinearKernel.h"
00016 #include "kernel/SparseKernel.h"
00017 
00018 CSparseLinearKernel::CSparseLinearKernel()
00019 : CSparseKernel<float64_t>(0), normal(NULL), normal_length(0)
00020 {
00021     properties |= KP_LINADD;
00022 }
00023 
00024 CSparseLinearKernel::CSparseLinearKernel(
00025     CSparseFeatures<float64_t>* l, CSparseFeatures<float64_t>* r)
00026 : CSparseKernel<float64_t>(0), normal(NULL), normal_length(0)
00027 {
00028     properties |= KP_LINADD;
00029     init(l,r);
00030 }
00031 
00032 CSparseLinearKernel::~CSparseLinearKernel()
00033 {
00034     cleanup();
00035 }
00036 
00037 bool CSparseLinearKernel::init(CFeatures* l, CFeatures* r)
00038 {
00039     CSparseKernel<float64_t>::init(l, r);
00040     return init_normalizer();
00041 }
00042 
00043 void CSparseLinearKernel::cleanup()
00044 {
00045     delete_optimization();
00046 
00047     CKernel::cleanup();
00048 }
00049 
00050 bool CSparseLinearKernel::load_init(FILE* src)
00051 {
00052     return false;
00053 }
00054 
00055 bool CSparseLinearKernel::save_init(FILE* dest)
00056 {
00057     return false;
00058 }
00059 
00060 void CSparseLinearKernel::clear_normal()
00061 {
00062     int32_t num=((CSparseFeatures<float64_t>*) lhs)->get_num_features();
00063     if (normal==NULL)
00064     {
00065         normal=new float64_t[num];
00066         normal_length=num;
00067     }
00068 
00069     memset(normal, 0, sizeof(float64_t)*normal_length);
00070     set_is_initialized(true);
00071 }
00072 
00073 void CSparseLinearKernel::add_to_normal(int32_t idx, float64_t weight)
00074 {
00075     ((CSparseFeatures<float64_t>*) rhs)->add_to_dense_vec(
00076         normalizer->normalize_lhs(weight, idx), idx, normal, normal_length);
00077     set_is_initialized(true);
00078 }
00079   
00080 float64_t CSparseLinearKernel::compute(int32_t idx_a, int32_t idx_b)
00081 {
00082   int32_t alen=0;
00083   int32_t blen=0;
00084   bool afree=false;
00085   bool bfree=false;
00086 
00087   TSparseEntry<float64_t>* avec=((CSparseFeatures<float64_t>*) lhs)->
00088     get_sparse_feature_vector(idx_a, alen, afree);
00089   TSparseEntry<float64_t>* bvec=((CSparseFeatures<float64_t>*) rhs)->
00090     get_sparse_feature_vector(idx_b, blen, bfree);
00091 
00092   float64_t result=((CSparseFeatures<float64_t>*) lhs)->
00093     sparse_dot(1.0, avec,alen, bvec,blen);
00094 
00095   ((CSparseFeatures<float64_t>*) lhs)->free_feature_vector(avec, idx_a, afree);
00096   ((CSparseFeatures<float64_t>*) rhs)->free_feature_vector(bvec, idx_b, bfree);
00097 
00098   return result;
00099 }
00100 
00101 bool CSparseLinearKernel::init_optimization(
00102     int32_t num_suppvec, int32_t* sv_idx, float64_t* alphas)
00103 {
00104     clear_normal();
00105 
00106     for (int32_t i=0; i<num_suppvec; i++)
00107         add_to_normal(sv_idx[i], alphas[i]);
00108 
00109     set_is_initialized(true);
00110     return true;
00111 }
00112 
00113 bool CSparseLinearKernel::delete_optimization()
00114 {
00115     delete[] normal;
00116     normal_length=0;
00117     normal=NULL;
00118     set_is_initialized(false);
00119 
00120     return true;
00121 }
00122 
00123 float64_t CSparseLinearKernel::compute_optimized(int32_t idx)
00124 {
00125     ASSERT(get_is_initialized());
00126     float64_t result = ((CSparseFeatures<float64_t>*) rhs)->
00127         dense_dot(1.0, idx, normal, normal_length, 0.0);
00128     return normalizer->normalize_rhs(result, idx);
00129 }

SHOGUN Machine Learning Toolbox - Documentation