Create an svm-problem of all available data. In case of FeatureTables all the tables in FeatureData are read to create one svm_problem. In case of distributed precomputed kernels it makes a dummy svm_problem. If Filter...Fold is called, this function wil only return a part of the whole set. The returned structure must be deleted with FreeProblem(). Implements Impala::Core::Training::TrainDataSrc. Definition at line 73 of file TrainDataSrcFeature.h. References Impala::Core::Training::SvmProblemBuilder::AddFeatureTable(), FeatureTableCount(), GetFeatureTableWithCache(), Impala::Core::Training::SvmProblemBuilder::MakeProblem(), Impala::Core::Training::TrainDataSrc::mAnnotation, mFeatureTableCache, and Impala::Core::Training::TrainDataSrc::mSelection. 00074 { 00075 SvmProblemBuilder pb(mAnnotation, mSelection); 00076 for(int i=0 ; i<FeatureTableCount() ; ++i) 00077 { 00078 Feature::FeatureTable* f = GetFeatureTableWithCache(i); 00079 pb.AddFeatureTable(f); 00080 } 00081 mFeatureTableCache.Empty(); 00082 return pb.MakeProblem(); 00083 }
Here is the call graph for this function:
|