diff -c ./v0/Common.cxx ./v2/Common.cxx *** ./v0/Common.cxx Tue Sep 5 11:22:32 2000 --- ./v2/Common.cxx Tue Sep 5 11:29:47 2000 *************** *** 23,26 **** const int Version = 3; const int Revision = 0; ! --- 23,26 ---- const int Version = 3; const int Revision = 0; ! const char *RevComment = ".2"; diff -c ./v0/Common.h ./v2/Common.h *** ./v0/Common.h Tue Sep 5 11:22:34 2000 --- ./v2/Common.h Tue Sep 5 11:29:48 2000 *************** *** 34,41 **** extern const int Version; extern const int Revision; - inline char look_ahead( istream &is ) { char ch; --- 34,41 ---- extern const int Version; extern const int Revision; + extern const char *RevComment; inline char look_ahead( istream &is ) { char ch; diff -c ./v0/GetOptClass.cxx ./v2/GetOptClass.cxx *** ./v0/GetOptClass.cxx Tue Sep 5 11:22:34 2000 --- ./v2/GetOptClass.cxx Tue Sep 5 11:29:48 2000 *************** *** 21,26 **** --- 21,30 ---- * Timbl@kub.nl */ + // + // 10 april 2000: Fixed a problem with get_N_option + // + #include #include #include *************** *** 337,344 **** if ( *Option++ == '-' ){ while( isspace(*Option) ) Option++; if ( *Option++ == 'N' ) { - Result = atoi( Option ); MaxFeatsSet = true; break; } } --- 341,355 ---- if ( *Option++ == '-' ){ while( isspace(*Option) ) Option++; if ( *Option++ == 'N' ) { MaxFeatsSet = true; + Result = 0; + if ( *Option ) + Result = atoi( Option ); + else if (opt < Argc ) + Result = atoi( Argv[opt] ); + if ( Result == 0 ){ + cerr << "missing a number after N option" << endl; + } break; } } diff -c ./v0/Instance.cxx ./v2/Instance.cxx *** ./v0/Instance.cxx Tue Sep 5 11:22:35 2000 --- ./v2/Instance.cxx Tue Sep 5 11:29:48 2000 *************** *** 19,24 **** --- 19,29 ---- * Timbl@kub.nl */ + // + // bug fixed 27-03-2000 Ko: ChiSquareStatistics: Resizing of n_i_dot was wrong + // 03-07-2000 Ko: NumStatistics: Test for ValFreq() >=0 should be + // a test for ValFreq() > 0. + // #include #include #include *************** *** 380,387 **** if ((map_iterator = Mapping.find( ID )) == Mapping.end()){ // Token was not found in map, so we insert a new copy of it here // ! Mapping.insert( IImaptype::value_type( ID, NumVals )); ! index = NumVals++; } else{ index = (*map_iterator).second; --- 385,392 ---- if ((map_iterator = Mapping.find( ID )) == Mapping.end()){ // Token was not found in map, so we insert a new copy of it here // ! Mapping.insert( IImaptype::value_type( ID, array_size )); ! index = array_size++; } else{ index = (*map_iterator).second; *************** *** 389,394 **** --- 394,418 ---- return index; } + int BaseFeatTargClass::localmaplookup( int ID ) + { + int index=-1; + IImaptype::iterator map_iterator; + if ((map_iterator = Mapping.find( ID )) != Mapping.end()){ + index = (*map_iterator).second; + } + return index; + } + + int BaseFeatTargClass::EffectiveValues() { + int result = 0; + int i; + for ( i=0; i < ArraySize(); i++ ) + if ( ValuesArray[i]->ValFreq() > 0 ) + result++; + return result; + } + FeatureValue *Feature::add_value( char *valstr, TargetValue *tv ) { HashInfo *Tok = TokenTree->Hash( valstr ); *************** *** 406,418 **** return (FeatureValue *)ValuesArray[index]; } BaseFeatTargClass::BaseFeatTargClass( int Size, int Inc, StringHash *T ) { CurSize = Size; Increment = Inc; TokenTree = T; ! Scraped = 0; ! NumVals = 0; TotalVals = 0; ValuesArray = (ValueClass **)malloc( sizeof(ValueClass*) * CurSize ); assert( ValuesArray != NULL ); --- 430,467 ---- return (FeatureValue *)ValuesArray[index]; } + bool Feature::increment_value( FeatureValue *FV, TargetValue *tv ) + { + bool result = false; + if ( FV ){ + TotalVals++; + FV->incr_val_freq(); + if ( tv ) + FV->TargetDist[tv] += 1; + result = true; + } + return result; + } + + bool Feature::decrement_value( FeatureValue *FV, TargetValue *tv ) + { + bool result = false; + if ( FV ){ + TotalVals--; + FV->decr_val_freq(); + if ( tv ) + FV->TargetDist[tv] -= 1; + result = true; + } + return result; + } + BaseFeatTargClass::BaseFeatTargClass( int Size, int Inc, StringHash *T ) { CurSize = Size; Increment = Inc; TokenTree = T; ! array_size = 0; TotalVals = 0; ValuesArray = (ValueClass **)malloc( sizeof(ValueClass*) * CurSize ); assert( ValuesArray != NULL ); *************** *** 470,480 **** bool Feature::prepare_numeric_stats( ) { ! double tmp, Total = 0.0; int freq; bool first = true; int i; ! for ( i=0; i < NumOfValues(); i++ ){ freq = ValuesArray[i]->ValFreq(); if ( freq ){ if ( !string2double( ValuesArray[i]->Name(), tmp ) ){ --- 519,529 ---- bool Feature::prepare_numeric_stats( ) { ! double tmp; int freq; bool first = true; int i; ! for ( i=0; i < ArraySize(); i++ ){ freq = ValuesArray[i]->ValFreq(); if ( freq ){ if ( !string2double( ValuesArray[i]->Name(), tmp ) ){ *************** *** 492,498 **** Min = tmp; else if ( tmp > Max ) Max = tmp; - Total += tmp * freq; } } } --- 541,546 ---- *************** *** 512,521 **** FVBin[i] = new FeatureValue( dumname ); } int j; ! for ( j=0; j < NumOfValues(); j++ ){ ! string2double( ValuesArray[j]->Name(), tmp ); ! place = (int)(floor((tmp - Min)/(Max-Min) * (BIN_SIZE-1)) ); ! FVBin[place]->TargetDist.Merge( &((FeatureValue *)ValuesArray[j])->TargetDist ); } int k; for ( k=0; k < BIN_SIZE; k++ ){ --- 560,571 ---- FVBin[i] = new FeatureValue( dumname ); } int j; ! for ( j=0; j < ArraySize(); j++ ){ ! if ( ValuesArray[j]->ValFreq() > 0 ){ ! string2double( ValuesArray[j]->Name(), tmp ); ! place = (int)(floor((tmp - Min)/(Max-Min) * (BIN_SIZE-1)) ); ! FVBin[place]->TargetDist.Merge( &((FeatureValue *)ValuesArray[j])->TargetDist ); ! } } int k; for ( k=0; k < BIN_SIZE; k++ ){ *************** *** 573,584 **** VDlist *hlp; FeatureValue **pnt = (FeatureValue**)ValuesArray; int i; ! for ( i=0; i < NumOfValues(); i++ ){ // Entropy for this FV pair. // double FVEntropy = 0.0; double Freq = pnt[i]->ValFreq(); ! if ( Freq ){ hlp = pnt[i]->TargetDist.First(); while ( hlp ){ Prob = hlp->Freq() / Freq; --- 623,634 ---- VDlist *hlp; FeatureValue **pnt = (FeatureValue**)ValuesArray; int i; ! for ( i=0; i < ArraySize(); i++ ){ // Entropy for this FV pair. // double FVEntropy = 0.0; double Freq = pnt[i]->ValFreq(); ! if ( Freq > 0 ){ hlp = pnt[i]->TargetDist.First(); while ( hlp ){ Prob = hlp->Freq() / Freq; *************** *** 599,607 **** SplitInfo = 0.0; pnt = (FeatureValue **)ValuesArray; int j; ! for ( j=0; j < NumOfValues(); j++ ){ Prob = pnt[j]->ValFreq() / (double)TotalValues(); ! if ( Prob ) SplitInfo += Prob * Log2(Prob); } SplitInfo = -SplitInfo; --- 649,657 ---- SplitInfo = 0.0; pnt = (FeatureValue **)ValuesArray; int j; ! for ( j=0; j < ArraySize(); j++ ){ Prob = pnt[j]->ValFreq() / (double)TotalValues(); ! if ( Prob > 0 ) SplitInfo += Prob * Log2(Prob); } SplitInfo = -SplitInfo; *************** *** 617,624 **** { VDlist *pnt; FeatureValue **FVA = (FeatureValue **)ValuesArray; ! int Size = Targets->NumOfValues(); ! int Num_Vals = NumOfValues(); long int n_dot_dot = 0; double tmp; if ( !n_dot_j ) { --- 667,674 ---- { VDlist *pnt; FeatureValue **FVA = (FeatureValue **)ValuesArray; ! int Size = Targets->ArraySize(); ! int Num_Vals = ArraySize(); long int n_dot_dot = 0; double tmp; if ( !n_dot_j ) { *************** *** 635,641 **** } if ( SaveNum < Num_Vals ){ delete [] n_i_dot; ! n_i_dot = new long int[Size]; SaveNum = Num_Vals; } } --- 685,691 ---- } if ( SaveNum < Num_Vals ){ delete [] n_i_dot; ! n_i_dot = new long int[Num_Vals]; SaveNum = Num_Vals; } } *************** *** 695,702 **** void Feature::SharedVarianceStatistics( Target *Targ ) { int NumInst = Targ->TotalValues(); ! int NumCats = Targ->NumOfValues(); ! int k = min( NumCats, NumOfValues() ) - 1 ; if ( k == 0 ) SharedVariance = 0; else --- 745,752 ---- void Feature::SharedVarianceStatistics( Target *Targ ) { int NumInst = Targ->TotalValues(); ! int NumCats = Targ->EffectiveValues(); ! int k = min( NumCats, EffectiveValues() ) - 1 ; if ( k == 0 ) SharedVariance = 0; else *************** *** 707,713 **** { if ( VDmatrix ){ int i; ! for ( i = 0; i < NumOfValues(); i++) free(VDmatrix[i]); free(VDmatrix); } --- 757,763 ---- { if ( VDmatrix ){ int i; ! for ( i = 0; i < ArraySize(); i++) free(VDmatrix[i]); free(VDmatrix); } *************** *** 722,738 **** { int i, j; if ( !Prestore_failed && VDmatrix == NULL ){ ! if ( NumOfValues() > 1000 ){ Prestore_failed = true; return false; } ! VDmatrix = (double**)malloc( NumOfValues() * sizeof(double*) ); if ( VDmatrix == NULL ){ Prestore_failed = true; return false; } ! for (i = 0; i < NumOfValues(); i++) { ! VDmatrix[i] = (double *)malloc( NumOfValues() * sizeof(double) ); if ( VDmatrix[i] == NULL ){ int si; for ( si = i-1; si >= 0; si-- ) --- 772,788 ---- { int i, j; if ( !Prestore_failed && VDmatrix == NULL ){ ! if ( ArraySize() > 1000 ){ Prestore_failed = true; return false; } ! VDmatrix = (double**)malloc( ArraySize() * sizeof(double*) ); if ( VDmatrix == NULL ){ Prestore_failed = true; return false; } ! for (i = 0; i < ArraySize(); i++) { ! VDmatrix[i] = (double *)malloc( ArraySize() * sizeof(double) ); if ( VDmatrix[i] == NULL ){ int si; for ( si = i-1; si >= 0; si-- ) *************** *** 745,752 **** } // Now we have the memory. Calculate and store. // ! for ( i = 0; i< NumOfValues(); i++ ){ ! for ( j=0; j < NumOfValues(); j++ ){ double result = 0.0; SA_list *p1 = ((FeatureValue *)ValuesArray[i])->SValueClassProb->First(); --- 795,802 ---- } // Now we have the memory. Calculate and store. // ! for ( i = 0; i< ArraySize(); i++ ){ ! for ( j=0; j < ArraySize(); j++ ){ double result = 0.0; SA_list *p1 = ((FeatureValue *)ValuesArray[i])->SValueClassProb->First(); *************** *** 785,791 **** { FeatureValue **FVA = (FeatureValue **)ValuesArray; int i; ! for ( i = 0; i < NumOfValues(); i++ ){ if ( FVA[i]->SValueClassProb ) os << FVA[i] << FVA[i]->SValueClassProb << endl; } --- 835,841 ---- { FeatureValue **FVA = (FeatureValue **)ValuesArray; int i; ! for ( i = 0; i < ArraySize(); i++ ){ if ( FVA[i]->SValueClassProb ) os << FVA[i] << FVA[i]->SValueClassProb << endl; } *************** *** 802,808 **** FeatureValue **FVA = (FeatureValue **)ValuesArray; // clear all existing arrays int i; ! for ( i = 0; i < NumOfValues(); i++ ){ if ( FVA[i]->SValueClassProb ){ delete FVA[i]->SValueClassProb; } --- 852,858 ---- FeatureValue **FVA = (FeatureValue **)ValuesArray; // clear all existing arrays int i; ! for ( i = 0; i < ArraySize(); i++ ){ if ( FVA[i]->SValueClassProb ){ delete FVA[i]->SValueClassProb; } *************** *** 853,859 **** } // check if we've got all the values, assign a default if not so int j; ! for ( j = 0; j < NumOfValues(); j++ ){ if ( FVA[j]->SValueClassProb == NULL ){ FVA[j]->SValueClassProb = new SparseArrayClass( Num ); FVA[j]->SValueClassProb->Assign( 0, -1.0 ); --- 903,909 ---- } // check if we've got all the values, assign a default if not so int j; ! for ( j = 0; j < ArraySize(); j++ ){ if ( FVA[j]->SValueClassProb == NULL ){ FVA[j]->SValueClassProb = new SparseArrayClass( Num ); FVA[j]->SValueClassProb->Assign( 0, -1.0 ); *************** *** 869,875 **** // { if ( shrt ){ ! cout << " a " << NumOfValues() << "x" << NumOfValues() << " matrix" << endl; } else { --- 919,925 ---- // { if ( shrt ){ ! cout << " a " << ArraySize() << "x" << ArraySize() << " matrix" << endl; } else { *************** *** 876,882 **** int old_prec = cout.precision(); cout.setf( ios::scientific ); int i; ! for ( i=0; i< NumOfValues(); i++ ){ cout.width(6); cout.setf(ios::left, ios::adjustfield); cout << (FeatureValue *)ValuesArray[i] << ":"; --- 926,932 ---- int old_prec = cout.precision(); cout.setf( ios::scientific ); int i; ! for ( i=0; i< ArraySize(); i++ ){ cout.width(6); cout.setf(ios::left, ios::adjustfield); cout << (FeatureValue *)ValuesArray[i] << ":"; *************** *** 884,890 **** cout.precision(3); cout.setf(ios::right, ios::adjustfield); int j; ! for ( j = 0; j < NumOfValues(); j++) { cout.width(12); cout.precision(3); cout.setf(ios::right, ios::adjustfield); --- 934,940 ---- cout.precision(3); cout.setf(ios::right, ios::adjustfield); int j; ! for ( j = 0; j < ArraySize(); j++) { cout.width(12); cout.precision(3); cout.setf(ios::right, ios::adjustfield); *************** *** 911,916 **** --- 961,988 ---- return (TargetValue *)ValuesArray[index]; } + bool Target::increment_value( TargetValue *TV ) + { + bool result = false; + if ( TV ){ + TotalVals++; + TV->incr_val_freq(); + result = true; + } + return result; + } + + bool Target::decrement_value( TargetValue *TV ) + { + bool result = false; + if ( TV ){ + TotalVals--; + TV->decr_val_freq(); + result = true; + } + return result; + } + Instance::Instance( int len ) { FV = new FeatureValue*[len]; diff -c ./v0/Instance.h ./v2/Instance.h *** ./v0/Instance.h Tue Sep 5 11:22:36 2000 --- ./v2/Instance.h Tue Sep 5 11:29:48 2000 *************** *** 226,246 **** ValueClass **ValuesArray; BaseFeatTargClass( int, int, StringHash * ); ~BaseFeatTargClass(); ! int NumOfValues() { return NumVals; }; ! unsigned int TotalValues() { return TotalVals - Scraped; }; ! int Scraped; int localmapping( int ); ValueClass *Lookup( char * ); protected: int CurSize; int Increment; unsigned int TotalVals; ! int NumVals; StringHash *TokenTree; IImaptype Mapping; void enlarge_values_array( int ); ! ValueClass *get_value_index( int I ) ! { return ValuesArray[I]; }; }; --- 226,246 ---- ValueClass **ValuesArray; BaseFeatTargClass( int, int, StringHash * ); ~BaseFeatTargClass(); ! int ArraySize() { return array_size; }; ! int EffectiveValues(); ! unsigned int TotalValues() { return TotalVals; }; int localmapping( int ); + int localmaplookup( int ); ValueClass *Lookup( char * ); protected: int CurSize; int Increment; unsigned int TotalVals; ! int array_size; StringHash *TokenTree; IImaptype Mapping; void enlarge_values_array( int ); ! ValueClass *get_value_index( int I ) { return ValuesArray[I]; }; }; *************** *** 248,253 **** --- 248,255 ---- public: Target( int a, int b, StringHash *T ): BaseFeatTargClass(a,b,T) {}; TargetValue *add_value( char * ); + bool decrement_value( TargetValue * ); + bool increment_value( TargetValue * ); }; class Feature: public BaseFeatTargClass { *************** *** 288,293 **** --- 290,297 ---- }; ~Feature(); FeatureValue *add_value( char *, TargetValue * ); + bool decrement_value( FeatureValue *, TargetValue * ); + bool increment_value( FeatureValue *, TargetValue * ); bool store_vd_matrix( ); void delete_vd_matrix( void ); void print_vd_matrix( bool s = false ); diff -c ./v0/MBLClass.cxx ./v2/MBLClass.cxx *** ./v0/MBLClass.cxx Tue Sep 5 11:22:36 2000 --- ./v2/MBLClass.cxx Tue Sep 5 11:29:48 2000 *************** *** 18,23 **** --- 18,30 ---- * Timbl@kub.nl */ + // + // 27-3-2000: Ko : UnHideInstance: Features[i] should be Permfeatures[i] + // The TargetDist must be increased as well + // HideInstance: The TargetDist should be decreased. + // 10-4-2000: Ko : Solved a problem with large output in write_perm + // + #include #include #include *************** *** 500,516 **** inline void MBLClass::write_perm(void) { int j; ! ostrstream outstr; ! outstr << "Feature Permutation based on " ! << order_to_string(TreeOrder, true) << " :\n"; ! outstr << "< "; for ( j=0; j < NumOfFeatures-1; j++ ){ ! outstr << Permutation[j]+1 << ", "; } ! outstr << Permutation[j]+1 << " >" << ends; ! char *pnt = outstr.str(); ! Info( pnt ); ! delete [] pnt; } inline void MBLClass::write_perm_special( ostream &os ) --- 507,519 ---- inline void MBLClass::write_perm(void) { int j; ! Info( "Feature Permutation based on %s :", ! order_to_string(TreeOrder, true) ); ! cerr << "< "; for ( j=0; j < NumOfFeatures-1; j++ ){ ! cerr << Permutation[j]+1 << ", "; } ! cerr << Permutation[j]+1 << " >" << endl; } inline void MBLClass::write_perm_special( ostream &os ) *************** *** 743,761 **** Order[i] = Features[i]->SharedVariance; break; case OneoverFeature: ! Order[i] = 1.0 / Features[i]->NumOfValues(); break; case GRoverFeature: ! Order[i] = Features[i]->GainRatio / Features[i]->NumOfValues(); break; case IGoverFeature: ! Order[i] = Features[i]->InfoGain / Features[i]->NumOfValues(); break; case X2overFeature: ! Order[i] = Features[i]->ChiSquare / Features[i]->NumOfValues(); break; case SVoverFeature: ! Order[i] = Features[i]->SharedVariance / Features[i]->NumOfValues(); break; case OneoverSplitInfo: Order[i] = 1.0 / Features[i]->SplitInfo; --- 746,764 ---- Order[i] = Features[i]->SharedVariance; break; case OneoverFeature: ! Order[i] = 1.0 / Features[i]->ArraySize(); break; case GRoverFeature: ! Order[i] = Features[i]->GainRatio / Features[i]->ArraySize(); break; case IGoverFeature: ! Order[i] = Features[i]->InfoGain / Features[i]->ArraySize(); break; case X2overFeature: ! Order[i] = Features[i]->ChiSquare / Features[i]->ArraySize(); break; case SVoverFeature: ! Order[i] = Features[i]->SharedVariance / Features[i]->ArraySize(); break; case OneoverSplitInfo: Order[i] = 1.0 / Features[i]->SplitInfo; *************** *** 973,982 **** // out << "Targets : "; int t; ! for ( t=0; t < Targets->NumOfValues()-1; t++ ) { out << tv[t] << ", "; } ! out << tv[Targets->NumOfValues()-1] << "." << endl << endl; int i; for ( i = 0; i < NumOfFeatures; i++ ) if ( Features[i]->Ignore ) --- 976,985 ---- // out << "Targets : "; int t; ! for ( t=0; t < Targets->ArraySize()-1; t++ ) { out << tv[t] << ", "; } ! out << tv[Targets->ArraySize()-1] << "." << endl << endl; int i; for ( i = 0; i < NumOfFeatures; i++ ) if ( Features[i]->Ignore ) *************** *** 997,1003 **** bool MBLClass::allocate_arrays() { ! int Dim = Targets->NumOfValues(); int i,j; for ( j = 0; j < NumOfFeatures; j++) { if ( !Features[j]->Ignore ) { --- 1000,1006 ---- bool MBLClass::allocate_arrays() { ! int Dim = Targets->ArraySize(); int i,j; for ( j = 0; j < NumOfFeatures; j++) { if ( !Features[j]->Ignore ) { *************** *** 1004,1010 **** FeatureValue **FVA = (FeatureValue **)Features[j]->ValuesArray; // Loop over all values. // ! for ( i=0; i < Features[j]->NumOfValues(); i++ ){ // Loop over all classes. if ( FVA[i]->SValueClassProb == NULL ){ if ( !(FVA[i]->SValueClassProb = --- 1007,1013 ---- FeatureValue **FVA = (FeatureValue **)Features[j]->ValuesArray; // Loop over all values. // ! for ( i=0; i < Features[j]->ArraySize(); i++ ){ // Loop over all classes. if ( FVA[i]->SValueClassProb == NULL ){ if ( !(FVA[i]->SValueClassProb = *************** *** 1034,1042 **** FeatureValue **FVA = (FeatureValue **)Features[j]->ValuesArray; // Loop over all values. // ! for ( i=0; i < Features[j]->NumOfValues(); i++ ){ double freq = FVA[i]->ValFreq(); ! if ( freq ){ // Loop over all classes. // VDlist *tmp = FVA[i]->TargetDist.First( ); --- 1037,1045 ---- FeatureValue **FVA = (FeatureValue **)Features[j]->ValuesArray; // Loop over all values. // ! for ( i=0; i < Features[j]->ArraySize(); i++ ){ double freq = FVA[i]->ValFreq(); ! if ( freq > 0 ){ // Loop over all classes. // VDlist *tmp = FVA[i]->TargetDist.First( ); *************** *** 1325,1331 **** for ( k = 0; k < EffectiveFeatures; k++ ){ register int j = Permutation[k]; CurrInst->FV[k] = (FeatureValue*)Features[j]->Lookup( OrgInput[j] ); ! } // i // and the Target CurrInst->TV = (TargetValue*)Targets->Lookup( OrgInput[NumOfFeatures] ); break; --- 1328,1334 ---- for ( k = 0; k < EffectiveFeatures; k++ ){ register int j = Permutation[k]; CurrInst->FV[k] = (FeatureValue*)Features[j]->Lookup( OrgInput[j] ); ! } // k // and the Target CurrInst->TV = (TargetValue*)Targets->Lookup( OrgInput[NumOfFeatures] ); break; *************** *** 1419,1425 **** s.precision(8); s.setf(ios::showpoint ); s << "DB Entropy : " << DBEntropy << endl; ! s << "Number of Classes : " << Targets->NumOfValues() << endl; s << endl; if ( verbosity & FEAT_W ){ s << "Feats\tVals\tX-square\tVariance\tInfoGain\tGainRatio" << endl; --- 1422,1429 ---- s.precision(8); s.setf(ios::showpoint ); s << "DB Entropy : " << DBEntropy << endl; ! // s << "Number of Classes : " << Targets->ArraySize() << endl; ! s << "Number of Classes : " << Targets->EffectiveValues() << endl; s << endl; if ( verbosity & FEAT_W ){ s << "Feats\tVals\tX-square\tVariance\tInfoGain\tGainRatio" << endl; *************** *** 1433,1439 **** else{ s.width(7); s.setf(ios::right, ios::adjustfield); ! s << Features[i]->NumOfValues() << "\t" << Features[i]->ChiSquare << "\t" << Features[i]->SharedVariance << "\t" << Features[i]->InfoGain --- 1437,1444 ---- else{ s.width(7); s.setf(ios::right, ios::adjustfield); ! // s << Features[i]->ArraySize() ! s << Features[i]->EffectiveValues() << "\t" << Features[i]->ChiSquare << "\t" << Features[i]->SharedVariance << "\t" << Features[i]->InfoGain *************** *** 1470,1476 **** if ( verbosity ) Info( "Saving Weights in %s", FileName ); outfile << "# DB Entropy: " << DBEntropy << endl; ! outfile << "# Classes: " << Targets->NumOfValues() << endl; outfile << "# Lines of data: " << Targets->TotalValues() << endl; outfile << "# Fea." << "\t" << "Weight" << endl; InitWeights( ); --- 1475,1481 ---- if ( verbosity ) Info( "Saving Weights in %s", FileName ); outfile << "# DB Entropy: " << DBEntropy << endl; ! outfile << "# Classes: " << Targets->ArraySize() << endl; outfile << "# Lines of data: " << Targets->TotalValues() << endl; outfile << "# Fea." << "\t" << "Weight" << endl; InitWeights( ); *************** *** 1839,1848 **** double Entropy = 0.0, Ratio; // first get the Database Entropy int i; ! for ( i=0; i < Targets->NumOfValues(); i++ ) { Ratio = Targets->ValuesArray[i]->ValFreq() / (double)Targets->TotalValues(); ! if ( Ratio ) Entropy += Ratio * Log2(Ratio); } DBEntropy = fabs(-Entropy); --- 1844,1853 ---- double Entropy = 0.0, Ratio; // first get the Database Entropy int i; ! for ( i=0; i < Targets->ArraySize(); i++ ) { Ratio = Targets->ValuesArray[i]->ValFreq() / (double)Targets->TotalValues(); ! if ( Ratio > 0 ) Entropy += Ratio * Log2(Ratio); } DBEntropy = fabs(-Entropy); *************** *** 1962,1971 **** // Print the possible classes. // int t; ! for ( t=0; t < Targets->NumOfValues()-1; t++ ) { namesfile << tv[t] << ","; } ! namesfile << tv[Targets->NumOfValues()-1] << "." << endl << endl; // Loop over the Features. // --- 1967,1976 ---- // Print the possible classes. // int t; ! for ( t=0; t < Targets->ArraySize()-1; t++ ) { namesfile << tv[t] << ","; } ! namesfile << tv[Targets->ArraySize()-1] << "." << endl << endl; // Loop over the Features. // *************** *** 1991,2000 **** // vf = (FeatureValue **)Features[f]->ValuesArray; int i; ! for ( i=0; i< Features[f]->NumOfValues()-1; i++ ){ namesfile << vf[i] << ","; } ! namesfile << vf[Features[f]->NumOfValues()-1] << "." << endl; } } --- 1996,2005 ---- // vf = (FeatureValue **)Features[f]->ValuesArray; int i; ! for ( i=0; i< Features[f]->ArraySize()-1; i++ ){ namesfile << vf[i] << ","; } ! namesfile << vf[Features[f]->ArraySize()-1] << "." << endl; } } *************** *** 2450,2470 **** InstanceBase->RemoveInstance( Inst ); int i; for ( i=0; i < EffectiveFeatures && result; i++ ){ ! if ( !Inst->FV[i] ){ FatalError( "Unable to Hide an Instance!" ); result = false; } - else { - Inst->FV[i]->decr_val_freq(); - PermFeatures[i]->delete_vd_matrix(); - PermFeatures[i]->Scraped = 1; - } } if ( result ){ ! Inst->TV->decr_val_freq(); ! Targets->Scraped = 1; ! MBLInit = false; ! result = prepare_MBL_test(); } return result; } --- 2455,2470 ---- InstanceBase->RemoveInstance( Inst ); int i; for ( i=0; i < EffectiveFeatures && result; i++ ){ ! if ( PermFeatures[i]->decrement_value( Inst->FV[i], Inst->TV ) ){ ! PermFeatures[i]->delete_vd_matrix(); ! } ! else { FatalError( "Unable to Hide an Instance!" ); result = false; } } if ( result ){ ! Targets->decrement_value( Inst->TV ); } return result; } *************** *** 2474,2484 **** InstanceBase->AddInstance( Inst ); int i; for ( i=0; i < EffectiveFeatures; i++ ){ ! Inst->FV[i]->incr_val_freq(); ! Features[i]->Scraped = 0; } ! Inst->TV->incr_val_freq(); ! Targets->Scraped = 0; } ValueDistribution *LOOClass::LocalClassify( int &status, --- 2474,2482 ---- InstanceBase->AddInstance( Inst ); int i; for ( i=0; i < EffectiveFeatures; i++ ){ ! PermFeatures[i]->increment_value( Inst->FV[i], Inst->TV ); } ! Targets->increment_value( Inst->TV ); } ValueDistribution *LOOClass::LocalClassify( int &status, *************** *** 2487,2493 **** status = -1; ValueDistribution *ResultDist = NULL; strings_to_instance( TestWords ); ! if ( HideInstance( CurrInst ) ){ if ( do_exact_match ) ResultDist = InstanceBase->ExactMatch( CurrInst ); if ( ResultDist ){ --- 2485,2492 ---- status = -1; ValueDistribution *ResultDist = NULL; strings_to_instance( TestWords ); ! MBLInit = false; ! if ( HideInstance( CurrInst ) && prepare_MBL_test() ){ if ( do_exact_match ) ResultDist = InstanceBase->ExactMatch( CurrInst ); if ( ResultDist ){ *************** *** 2500,2508 **** ResultDist = test_instance_full( CurrInst, Distance ); } UnHideInstance( CurrInst ); } else { ! Warning( " Hiding failed!\n" ); } return ResultDist; } --- 2499,2508 ---- ResultDist = test_instance_full( CurrInst, Distance ); } UnHideInstance( CurrInst ); + MBLInit = false; } else { ! FatalError( " Hiding failed!\n" ); } return ResultDist; } diff -c ./v0/Makefile ./v2/Makefile *** ./v0/Makefile Tue Sep 5 11:22:37 2000 --- ./v2/Makefile Tue Sep 5 11:29:48 2000 *************** *** 9,14 **** --- 9,21 ---- OBJS = $(SRCS:.cxx=.o) EXES = $(MSRCS:.cxx=) + #INSTALL_DIR = /usr/local/bin + #LIB_INSTALL_DIR = /usr/local/lib + INSTALL_DIR = ./SPARC + LIB_INSTALL_DIR = ./SPARC + #INSTALL_DIR = ./X86 + #LIB_INSTALL_DIR = ./X86 + all: -@if [ $(OSTYPE) = linux ] ||\ [ $(OSTYPE) = Linux ] || \ *************** *** 44,48 **** -rm *.tmp install: ! cp $(EXES) /usr/local/bin ! cp libTimbl.a /usr/local/lib --- 51,55 ---- -rm *.tmp install: ! cp $(EXES) $(INSTALL_DIR) ! cp libTimbl.a $(LIB_INSTALL_DIR) Only in ./v0: SPARC diff -c ./v0/Timbl.cxx ./v2/Timbl.cxx *** ./v0/Timbl.cxx Tue Sep 5 11:22:38 2000 --- ./v2/Timbl.cxx Tue Sep 5 11:29:49 2000 *************** *** 47,53 **** string TheTestFile, PercFile; // Start. // ! cerr << "TiMBL " << Version << "." << Revision << " (c) ILK 1998, 1999, 2000.\n" << "Tilburg Memory Based Learner\n" << "Induction of Linguistic Knowledge Research Group\n" --- 47,53 ---- string TheTestFile, PercFile; // Start. // ! cerr << "TiMBL " << Version << "." << Revision << RevComment << " (c) ILK 1998, 1999, 2000.\n" << "Tilburg Memory Based Learner\n" << "Induction of Linguistic Knowledge Research Group\n" Only in ./v0: X86 Only in ./v2: patch301-302