SHOGUN
v3.2.0
|
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) 1999-2009 Soeren Sonnenburg 00008 * Copyright (C) 1999-2009 Fraunhofer Institute FIRST and Max-Planck-Society 00009 */ 00010 00011 #ifndef _ONLINELINEARCLASSIFIER_H__ 00012 #define _ONLINELINEARCLASSIFIER_H__ 00013 00014 #include <shogun/lib/common.h> 00015 #include <shogun/labels/Labels.h> 00016 #include <shogun/labels/RegressionLabels.h> 00017 #include <shogun/features/streaming/StreamingDotFeatures.h> 00018 #include <shogun/machine/Machine.h> 00019 00020 #include <stdio.h> 00021 00022 namespace shogun 00023 { 00050 class COnlineLinearMachine : public CMachine 00051 { 00052 public: 00054 COnlineLinearMachine(); 00055 virtual ~COnlineLinearMachine(); 00056 00062 virtual void get_w(float32_t*& dst_w, int32_t& dst_dims) 00063 { 00064 ASSERT(w && w_dim>0) 00065 dst_w=w; 00066 dst_dims=w_dim; 00067 } 00068 00075 virtual void get_w(float64_t*& dst_w, int32_t& dst_dims) 00076 { 00077 ASSERT(w && w_dim>0) 00078 dst_w=SG_MALLOC(float64_t, w_dim); 00079 for (int32_t i=0; i<w_dim; i++) 00080 dst_w[i]=w[i]; 00081 dst_dims=w_dim; 00082 } 00083 00088 virtual SGVector<float32_t> get_w() 00089 { 00090 float32_t * dst_w = SG_MALLOC(float32_t, w_dim); 00091 for (int32_t i=0; i<w_dim; i++) 00092 dst_w[i]=w[i]; 00093 return SGVector<float32_t>(dst_w, w_dim); 00094 } 00095 00101 virtual void set_w(float32_t* src_w, int32_t src_w_dim) 00102 { 00103 SG_FREE(w); 00104 w=SG_MALLOC(float32_t, src_w_dim); 00105 memcpy(w, src_w, size_t(src_w_dim)*sizeof(float32_t)); 00106 w_dim=src_w_dim; 00107 } 00108 00115 virtual void set_w(float64_t* src_w, int32_t src_w_dim) 00116 { 00117 SG_FREE(w); 00118 w=SG_MALLOC(float32_t, src_w_dim); 00119 for (int32_t i=0; i<src_w_dim; i++) 00120 w[i] = src_w[i]; 00121 w_dim=src_w_dim; 00122 } 00123 00128 virtual void set_bias(float32_t b) 00129 { 00130 bias=b; 00131 } 00132 00137 virtual float32_t get_bias() 00138 { 00139 return bias; 00140 } 00141 00146 virtual void set_features(CStreamingDotFeatures* feat) 00147 { 00148 SG_REF(feat); 00149 SG_UNREF(features); 00150 features=feat; 00151 } 00152 00159 virtual CRegressionLabels* apply_regression(CFeatures* data=NULL); 00160 00167 virtual CBinaryLabels* apply_binary(CFeatures* data=NULL); 00168 00170 virtual float64_t apply_one(int32_t vec_idx) 00171 { 00172 SG_NOTIMPLEMENTED 00173 return CMath::INFTY; 00174 } 00175 00184 virtual float32_t apply_one(float32_t* vec, int32_t len); 00185 00191 virtual float32_t apply_to_current_example(); 00192 00197 virtual CStreamingDotFeatures* get_features() { SG_REF(features); return features; } 00198 00204 virtual const char* get_name() const { return "OnlineLinearMachine"; } 00205 00209 virtual void start_train() { } 00210 00214 virtual void stop_train() { } 00215 00225 virtual void train_example(CStreamingDotFeatures *feature, float64_t label) { SG_NOTIMPLEMENTED } 00226 00227 protected: 00236 virtual bool train_machine(CFeatures* data=NULL); 00237 00243 SGVector<float64_t> apply_get_outputs(CFeatures* data); 00244 00246 virtual bool train_require_labels() const { return false; } 00247 00248 protected: 00250 int32_t w_dim; 00252 float32_t* w; 00254 float32_t bias; 00256 CStreamingDotFeatures* features; 00257 }; 00258 } 00259 #endif