Go to the documentation of this file.00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011 #include <shogun/lib/config.h>
00012
00013 #ifdef USE_CPLEX
00014
00015 #include <shogun/classifier/LPBoost.h>
00016 #include <shogun/labels/Labels.h>
00017 #include <shogun/mathematics/Math.h>
00018 #include <shogun/mathematics/Cplex.h>
00019 #include <shogun/lib/DynamicArray.h>
00020 #include <shogun/lib/Signal.h>
00021 #include <shogun/lib/Time.h>
00022
00023 using namespace shogun;
00024
00025 CLPBoost::CLPBoost()
00026 : CLinearMachine(), C1(1), C2(1), use_bias(true), epsilon(1e-3)
00027 {
00028 u=NULL;
00029 dim=NULL;
00030 num_sfeat=0;
00031 num_svec=0;
00032 sfeat=NULL;
00033 }
00034
00035
00036 CLPBoost::~CLPBoost()
00037 {
00038 cleanup();
00039 }
00040
00041 bool CLPBoost::init(int32_t num_vec)
00042 {
00043 u=SG_MALLOC(float64_t, num_vec);
00044 for (int32_t i=0; i<num_vec; i++)
00045 u[i]=1.0/num_vec;
00046
00047 dim=new CDynamicArray<int32_t>(100000);
00048
00049 sfeat= ((CSparseFeatures<float64_t>*) features)->get_transposed(num_sfeat, num_svec);
00050
00051 if (sfeat)
00052 return true;
00053 else
00054 return false;
00055 }
00056
00057 void CLPBoost::cleanup()
00058 {
00059 SG_FREE(u);
00060 u=NULL;
00061
00062 ((CSparseFeatures<float64_t>*) features)->clean_tsparse(sfeat, num_svec);
00063 sfeat=NULL;
00064
00065 delete dim;
00066 dim=NULL;
00067 }
00068
00069 float64_t CLPBoost::find_max_violator(int32_t& max_dim)
00070 {
00071 float64_t max_val=0;
00072 max_dim=-1;
00073
00074 for (int32_t i=0; i<num_svec; i++)
00075 {
00076 float64_t valplus=0;
00077 float64_t valminus=0;
00078
00079 for (int32_t j=0; j<sfeat[i].num_feat_entries; j++)
00080 {
00081 int32_t idx=sfeat[i].features[j].feat_index;
00082 float64_t v=u[idx]*((CBinaryLabels*)m_labels)->get_confidence(idx)*sfeat[i].features[j].entry;
00083 valplus+=v;
00084 valminus-=v;
00085 }
00086
00087 if (valplus>max_val || max_dim==-1)
00088 {
00089 max_dim=i;
00090 max_val=valplus;
00091 }
00092
00093 if (valminus>max_val)
00094 {
00095 max_dim=num_svec+i;
00096 max_val=valminus;
00097 }
00098 }
00099
00100 dim->append_element(max_dim);
00101 return max_val;
00102 }
00103
00104 bool CLPBoost::train_machine(CFeatures* data)
00105 {
00106 ASSERT(m_labels);
00107 ASSERT(features);
00108 int32_t num_train_labels=m_labels->get_num_labels();
00109 int32_t num_feat=features->get_dim_feature_space();
00110 int32_t num_vec=features->get_num_vectors();
00111
00112 ASSERT(num_vec==num_train_labels);
00113 w = SGVector<float64_t>(num_feat);
00114 memset(w.vector,0,sizeof(float64_t)*num_feat);
00115
00116 CCplex solver;
00117 solver.init(E_LINEAR);
00118 SG_PRINT("setting up lpboost\n");
00119 solver.setup_lpboost(C1, num_vec);
00120 SG_PRINT("finished setting up lpboost\n");
00121
00122 float64_t result=init(num_vec);
00123 ASSERT(result);
00124
00125 int32_t num_hypothesis=0;
00126 CTime time;
00127 CSignal::clear_cancel();
00128
00129 while (!(CSignal::cancel_computations()))
00130 {
00131 int32_t max_dim=0;
00132 float64_t violator=find_max_violator(max_dim);
00133 SG_PRINT("iteration:%06d violator: %10.17f (>1.0) chosen: %d\n", num_hypothesis, violator, max_dim);
00134 if (violator <= 1.0+epsilon && num_hypothesis>1)
00135 {
00136 SG_PRINT("converged after %d iterations!\n", num_hypothesis);
00137 break;
00138 }
00139
00140 float64_t factor=+1.0;
00141 if (max_dim>=num_svec)
00142 {
00143 factor=-1.0;
00144 max_dim-=num_svec;
00145 }
00146
00147 SGSparseVectorEntry<float64_t>* h=sfeat[max_dim].features;
00148 int32_t len=sfeat[max_dim].num_feat_entries;
00149 solver.add_lpboost_constraint(factor, h, len, num_vec, m_labels);
00150 solver.optimize(u);
00151
00152 num_hypothesis++;
00153
00154 if (get_max_train_time()>0 && time.cur_time_diff()>get_max_train_time())
00155 break;
00156 }
00157 float64_t* lambda=SG_MALLOC(float64_t, num_hypothesis);
00158 solver.optimize(u, lambda);
00159
00160
00161 for (int32_t i=0; i<num_hypothesis; i++)
00162 {
00163 int32_t d=dim->get_element(i);
00164 if (d>=num_svec)
00165 w[d-num_svec]+=lambda[i];
00166 else
00167 w[d]-=lambda[i];
00168
00169 }
00170
00171 solver.cleanup();
00172
00173 cleanup();
00174
00175 return true;
00176 }
00177 #endif