diff -c v0/Common.cxx v3/Common.cxx *** v0/Common.cxx Tue Sep 5 11:22:32 2000 --- v3/Common.cxx Wed Sep 20 16:44:38 2000 *************** *** 23,26 **** const int Version = 3; const int Revision = 0; ! --- 23,26 ---- const int Version = 3; const int Revision = 0; ! const char *RevComment = ".3"; diff -c v0/Common.h v3/Common.h *** v0/Common.h Tue Sep 5 11:22:34 2000 --- v3/Common.h Wed Sep 20 16:44:45 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 v3/GetOptClass.cxx *** v0/GetOptClass.cxx Tue Sep 5 11:22:34 2000 --- v3/GetOptClass.cxx Wed Sep 20 16:44:38 2000 *************** *** 21,26 **** --- 21,30 ---- * Timbl@kub.nl */ + // + // 10 april 2000: Fixed a problem with get_N_option + // 23-08-200 : free the result of verbosity_to_string + #include #include #include *************** *** 180,186 **** } if ( TestFileName != "" ){ cerr << "Multiple testfile name " << TestFileName << " ignored.\n" ! << "since -f leave_one_out is also specified" << endl; } if ( LocalAlgo != IB1 ) { cerr << "option -a ignored, since it conflicts with -t leave_one_out\n" --- 184,190 ---- } if ( TestFileName != "" ){ cerr << "Multiple testfile name " << TestFileName << " ignored.\n" ! << "since -t leave_one_out is also specified" << endl; } if ( LocalAlgo != IB1 ) { cerr << "option -a ignored, since it conflicts with -t leave_one_out\n" *************** *** 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; } } *************** *** 462,474 **** inline void show_version_info(void) { ! cerr << "TiMBL " << Version << "." << Revision << ", compiled on " << __DATE__ << ", " << __TIME__ << endl; exit(1); } bool GetOptClass::definitive_options( TimblExperiment *Exp, ! bool first ){ char optline[3 * OPT_LENGTH]; if ( !sparse_bin_set && !to_late_for_bin && (LocalInputFormat != UnknownInputFormat || LocalAlgo != LOO ) ){ --- 473,485 ---- inline void show_version_info(void) { ! cerr << "TiMBL " << Version << "." << Revision << RevComment << ", compiled on " << __DATE__ << ", " << __TIME__ << endl; exit(1); } bool GetOptClass::definitive_options( TimblExperiment *Exp, ! bool first ){ char optline[3 * OPT_LENGTH]; if ( !sparse_bin_set && !to_late_for_bin && (LocalInputFormat != UnknownInputFormat || LocalAlgo != LOO ) ){ *************** *** 522,529 **** if ( Exp->SetOption( optline ) ){ sprintf( optline, "PROGRESS: %d", local_progress ); if ( Exp->SetOption( optline ) ){ ! sprintf( optline, "VERBOSITY: %s", ! verbosity_to_string(MyVerbosity) ); if ( Exp->SetOption( optline ) ){ int i; for ( i=0; i < MaxFeats+1; i++ ){ --- 533,541 ---- if ( Exp->SetOption( optline ) ){ sprintf( optline, "PROGRESS: %d", local_progress ); if ( Exp->SetOption( optline ) ){ ! const char *tmp = verbosity_to_string(MyVerbosity); ! sprintf( optline, "VERBOSITY: %s", tmp ); ! free( const_cast(tmp) ); if ( Exp->SetOption( optline ) ){ int i; for ( i=0; i < MaxFeats+1; i++ ){ *************** *** 532,542 **** if (!Exp->SetOption( optline ) ) return false; } ! if ( do_query ){ ! Exp->ShowSettings( ); ! do_query = false; } - return true; } } } --- 544,558 ---- if (!Exp->SetOption( optline ) ) return false; } ! sprintf( optline, "TREE_ORDER: %s", ! order_to_string(LocalOrder) ); ! if ( Exp->SetOption( optline ) ){ ! if ( do_query ){ ! Exp->ShowSettings( ); ! do_query = false; ! } ! return true; } } } } diff -c v0/IBtree.cxx v3/IBtree.cxx *** v0/IBtree.cxx Tue Sep 5 10:51:21 2000 --- v3/IBtree.cxx Wed Sep 20 16:44:38 2000 *************** *** 19,24 **** --- 19,30 ---- * Timbl@kub.nl */ + // bug fixed 23-08-2000 Ko: replaced "magic number" 80 with MAX_BUFFER_SIZE + // in read_local and read_list + // 20-09-2000 Ko: Changed read_local. Shouldn't add targetvalues but + // look them up! (they are added by read_distribution) + // + #include #include #include *************** *** 206,212 **** { IBtree *result = new IBtree; tree_size++; ! char buf[80]; is >> buf; result->FValue = Feats[level]->add_value( buf, NULL ); is >> buf; --- 212,218 ---- { IBtree *result = new IBtree; tree_size++; ! char buf[MAX_BUFFER_SIZE]; is >> buf; result->FValue = Feats[level]->add_value( buf, NULL ); is >> buf; *************** *** 216,229 **** return NULL; } is >> buf; ! result->TValue = Targ->add_value( buf ); if ( look_ahead(is) == '{' ){ result->TDistribution = new ValueDistribution(); ! result->TDistribution->read_distribution( is, Targ, false ); // cerr << "Dist = " << TopDistribution << endl; // also we have to update the targetinformation of the featurevalue // so we can recalculate the statistics later on. ! result->FValue->ReconstructDistribution( result->TDistribution ); } if ( look_ahead(is) == '[' ){ result->link = read_list( is, Feats, Targ, level+1 ); --- 222,237 ---- return NULL; } is >> buf; ! result->TValue = (TargetValue*)Targ->Lookup( buf ); if ( look_ahead(is) == '{' ){ result->TDistribution = new ValueDistribution(); ! if ( !result->TDistribution->read_distribution( is, Targ, false ) ) ! Error( "problems reading a distribution from InstanceBase file" ); // cerr << "Dist = " << TopDistribution << endl; // also we have to update the targetinformation of the featurevalue // so we can recalculate the statistics later on. ! if ( result->FValue->ValFreq() > 0 ) ! result->FValue->ReconstructDistribution( result->TDistribution ); } if ( look_ahead(is) == '[' ){ result->link = read_list( is, Feats, Targ, level+1 ); *************** *** 256,262 **** IBtree *result = NULL, **pnt; pnt = &result; bool goon = true; ! char buf[80]; if ( Version < 3 ) is >> buf; // skip the opening `[` (old style) while ( goon ) { --- 264,270 ---- IBtree *result = NULL, **pnt; pnt = &result; bool goon = true; ! char buf[MAX_BUFFER_SIZE]; if ( Version < 3 ) is >> buf; // skip the opening `[` (old style) while ( goon ) { *************** *** 300,306 **** is >> buf; // Now read the TopDistribution, to get the Targets // in the right order in Targ ! TopDistribution->read_distribution( is, Targs, true ); // cerr << "Top = " << TopDistribution << endl; if ( look_ahead( is ) == '[' ){ InstBase = read_list( is, Feats, Targs, 0 ); --- 308,315 ---- is >> buf; // Now read the TopDistribution, to get the Targets // in the right order in Targ ! if ( !TopDistribution->read_distribution( is, Targs, true ) ) ! Error( "problems reading Top Distribution " ); // cerr << "Top = " << TopDistribution << endl; if ( look_ahead( is ) == '[' ){ InstBase = read_list( is, Feats, Targs, 0 ); *************** *** 506,512 **** result = false; } else if ( !InvAss ){ - // IBtree *Arr[Depth]; IBtree **Arr = new IBtree *[Depth+1]; IBtree *pnt = InstBase; int CurI = 0, Feat = 0; --- 515,520 ---- diff -c v0/Instance.cxx v3/Instance.cxx *** v0/Instance.cxx Tue Sep 5 11:22:35 2000 --- v3/Instance.cxx Wed Sep 20 16:44:39 2000 *************** *** 19,24 **** --- 19,31 ---- * 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. + // 20-09-2000 Ko: Repaired problem with Target::add_value, which made + // TotalValues() useless. Also affected: read_distribution + // #include #include #include *************** *** 41,48 **** #define BIN_SIZE 20 ! ostream& operator<<(ostream& s, const FeatureValue* fv ) ! { if ( fv ) s << fv->TokVal->Name(); else --- 48,54 ---- #define BIN_SIZE 20 ! ostream& operator<<(ostream& s, const FeatureValue* fv ){ if ( fv ) s << fv->TokVal->Name(); else *************** *** 50,57 **** return s; } ! ostream& operator<<(ostream& s, const TargetValue* tv ) ! { if ( tv ) s << tv->TokVal->Name(); else --- 56,62 ---- return s; } ! ostream& operator<<(ostream& s, const TargetValue* tv ){ if ( tv ) s << tv->TokVal->Name(); else *************** *** 230,257 **** } } ! void ValueDistribution::read_distribution( istream &is, Target *Targ, ! bool do_fr ) ! { // read a distribution from stream is into Target // if do_f we also adjust the value of Frequency of the Target, which is // otherwise 1. Special case when reading the TopDistribution. // TargetValue *target; char buf[STR_LEN]; int freq; is >> buf; // skip { do { is >> buf; ! target = Targ->add_value( buf ); ! is >> buf; ! freq = atoi( buf ); ! if ( do_fr ) ! target->ValFreq( freq ); (*this)[target] = freq; } while ( look_ahead(is) != '}' ); is >> buf; // skip } } --- 235,270 ---- } } ! bool ValueDistribution::read_distribution( istream &is, Target *Targ, ! bool do_fr ){ // read a distribution from stream is into Target // if do_f we also adjust the value of Frequency of the Target, which is // otherwise 1. Special case when reading the TopDistribution. // + bool result = true; TargetValue *target; char buf[STR_LEN]; + char buf2[STR_LEN]; int freq; is >> buf; // skip { do { is >> buf; ! is >> buf2; ! freq = atoi( buf2 ); ! if ( do_fr ){ ! target = Targ->add_value( buf, freq ); ! } ! else ! target = (TargetValue*)Targ->Lookup( buf ); ! if ( !target ){ ! result = false; ! break; ! } (*this)[target] = freq; } while ( look_ahead(is) != '}' ); is >> buf; // skip } + return result; } *************** *** 263,269 **** double randnum = (double)rand()/(double)RAND_MAX; randnum *= (Max-Min); randnum += Min; ! return (int)rint(randnum); } const TargetValue *ValueDistribution::BestTarget( bool& tie, --- 276,282 ---- double randnum = (double)rand()/(double)RAND_MAX; randnum *= (Max-Min); randnum += Min; ! return (int)floor(randnum+0.5); } const TargetValue *ValueDistribution::BestTarget( bool& tie, *************** *** 357,364 **** index = I; } ! FeatureValue::FeatureValue( char *s ) ! { SValueClassProb = NULL; Frequency = 0; index = -1; --- 370,376 ---- index = I; } ! FeatureValue::FeatureValue( char *s ){ SValueClassProb = NULL; Frequency = 0; index = -1; *************** *** 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; --- 392,399 ---- 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; *************** *** 389,399 **** return index; } ! FeatureValue *Feature::add_value( char *valstr, TargetValue *tv ) ! { HashInfo *Tok = TokenTree->Hash( valstr ); int index = localmapping( Tok->Index() ); - TotalVals++; if ( index >= CurSize ) enlarge_values_array( index ); if( ValuesArray[index] == NULL ){ --- 401,427 ---- return index; } ! int BaseFeatTargClass::EffectiveValues() { ! int result = 0; ! int i; ! for ( i=0; i < ArraySize(); i++ ) ! if ( ValuesArray[i]->ValFreq() > 0 ) ! result++; ! return result; ! } ! ! unsigned int BaseFeatTargClass::TotalValues() { ! unsigned int result = 0; ! int i; ! for ( i=0; i < ArraySize(); i++ ) ! result += ValuesArray[i]->ValFreq(); ! return result; ! } ! ! FeatureValue *Feature::add_value( char *valstr, ! TargetValue *tv ){ HashInfo *Tok = TokenTree->Hash( valstr ); int index = localmapping( Tok->Index() ); if ( index >= CurSize ) enlarge_values_array( index ); if( ValuesArray[index] == NULL ){ *************** *** 406,419 **** 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 ); int i; --- 434,469 ---- return (FeatureValue *)ValuesArray[index]; } + bool Feature::increment_value( FeatureValue *FV, TargetValue *tv ) + { + bool result = false; + if ( FV ){ + 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 ){ + 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; ValuesArray = (ValueClass **)malloc( sizeof(ValueClass*) * CurSize ); assert( ValuesArray != NULL ); int i; *************** *** 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 ) ){ --- 520,530 ---- 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; } } } --- 542,547 ---- *************** *** 503,508 **** --- 552,558 ---- { int place; double tmp, Prob, FVEntropy, Freq; + unsigned int TotalVals = TotalValues(); char dumname[STR_LEN]; FeatureValue **FVBin = new FeatureValue *[BIN_SIZE]; Entropy = 0.0; *************** *** 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++ ){ --- 562,573 ---- 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++ ){ *************** *** 532,538 **** FVEntropy += Prob * Log2(Prob); hlp = hlp->Next(); } ! Entropy += -FVEntropy * Freq / (double)TotalValues(); } } Entropy = fabs( Entropy ); --- 584,590 ---- FVEntropy += Prob * Log2(Prob); hlp = hlp->Next(); } ! Entropy += -FVEntropy * Freq / (double)TotalVals; } } Entropy = fabs( Entropy ); *************** *** 547,553 **** for ( m=0; m < BIN_SIZE; m++ ){ Freq = ((FeatureValue*)FVBin[m])->ValFreq(); if ( Freq ){ ! Prob = Freq / (double)TotalValues(); SplitInfo += Prob * Log2(Prob); } delete (FeatureValue*)FVBin[m]; --- 599,605 ---- for ( m=0; m < BIN_SIZE; m++ ){ Freq = ((FeatureValue*)FVBin[m])->ValFreq(); if ( Freq ){ ! Prob = Freq / (double)TotalVals; SplitInfo += Prob * Log2(Prob); } delete (FeatureValue*)FVBin[m]; *************** *** 568,584 **** void Feature::Statistics( double DBentropy ) { double Prob = 0.0; Entropy = 0.0; // Loop over the values. 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; --- 620,637 ---- void Feature::Statistics( double DBentropy ) { double Prob = 0.0; + unsigned int TotalVals = TotalValues(); Entropy = 0.0; // Loop over the values. 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; *************** *** 585,591 **** FVEntropy += Prob * Log2(Prob); hlp = hlp->Next(); } ! Entropy += -FVEntropy * Freq / (double)TotalValues(); } } //end i --- 638,644 ---- FVEntropy += Prob * Log2(Prob); hlp = hlp->Next(); } ! Entropy += -FVEntropy * Freq / (double)TotalVals; } } //end i *************** *** 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; --- 652,660 ---- SplitInfo = 0.0; pnt = (FeatureValue **)ValuesArray; int j; ! for ( j=0; j < ArraySize(); j++ ){ ! Prob = pnt[j]->ValFreq() / (double)TotalVals; ! 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 ) { --- 670,677 ---- { 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; } } --- 688,694 ---- } 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 --- 748,755 ---- 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); } --- 760,766 ---- { 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-- ) --- 775,791 ---- { 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(); --- 798,805 ---- } // 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; } --- 838,844 ---- { 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; } --- 855,861 ---- 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 ); --- 906,912 ---- } // 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 { --- 922,928 ---- // { 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] << ":"; --- 929,935 ---- 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); --- 937,943 ---- 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); *************** *** 896,916 **** } } ! TargetValue *Target::add_value( char *valstr ) ! { HashInfo *Tok = TokenTree->Hash( valstr ); int index = localmapping( Tok->Index() ); - TotalVals++; if ( index >= CurSize ) enlarge_values_array( index ); if( ValuesArray[index] == NULL ){ ValuesArray[index] = new TargetValue( Tok, index ); } else ! ValuesArray[index]->incr_val_freq(); return (TargetValue *)ValuesArray[index]; } Instance::Instance( int len ) { FV = new FeatureValue*[len]; --- 949,988 ---- } } ! TargetValue *Target::add_value( char *valstr, int freq ){ HashInfo *Tok = TokenTree->Hash( valstr ); int index = localmapping( Tok->Index() ); if ( index >= CurSize ) enlarge_values_array( index ); if( ValuesArray[index] == NULL ){ ValuesArray[index] = new TargetValue( Tok, index ); + ValuesArray[index]->ValFreq( freq ); } else ! ValuesArray[index]->ValFreq( ValuesArray[index]->ValFreq() + freq ); return (TargetValue *)ValuesArray[index]; } + bool Target::increment_value( TargetValue *TV ) + { + bool result = false; + if ( TV ){ + TV->incr_val_freq(); + result = true; + } + return result; + } + + bool Target::decrement_value( TargetValue *TV ) + { + bool result = false; + if ( TV ){ + TV->decr_val_freq(); + result = true; + } + return result; + } + Instance::Instance( int len ) { FV = new FeatureValue*[len]; diff -c v0/Instance.h v3/Instance.h *** v0/Instance.h Tue Sep 5 11:22:36 2000 --- v3/Instance.h Wed Sep 20 16:44:45 2000 *************** *** 20,26 **** #define INSTANCE_H ! #define OWN_MEM_HANDLING YES #include "Stack.h" --- 20,26 ---- #define INSTANCE_H ! // #define OWN_MEM_HANDLING YES #include "Stack.h" *************** *** 77,83 **** const TargetValue* BestTarget( ) const; void Merge( ValueDistribution *vd ); void Clear(){ delete distribution; distribution = NULL; }; ! void read_distribution( istream &, Target *, bool ); void DistToString( string& ) const; bool ZeroDist() { return ( distribution == 0 || --- 77,83 ---- const TargetValue* BestTarget( ) const; void Merge( ValueDistribution *vd ); void Clear(){ delete distribution; distribution = NULL; }; ! bool read_distribution( istream &, Target *, bool ); void DistToString( string& ) const; bool ZeroDist() { return ( distribution == 0 || *************** *** 150,160 **** template void SparseArrayClass::Assign( const int I, ! const ContentType val ) // search for I, if not there: add entry // assigns prob with val; // adjusts Dimension when necessary - { SA_list *tmp, **pnt = &contents; while ( *pnt ){ if ( I < (*pnt)->index ){ --- 150,159 ---- template void SparseArrayClass::Assign( const int I, ! const ContentType val ){ // search for I, if not there: add entry // assigns prob with val; // adjusts Dimension when necessary SA_list *tmp, **pnt = &contents; while ( *pnt ){ if ( I < (*pnt)->index ){ *************** *** 178,185 **** }; template ostream& operator<< ! (ostream& os, SparseArrayClass* SArr ) ! { if ( SArr ) { int old_prec = os.precision(); os.precision(3); --- 177,183 ---- }; template ostream& operator<< ! (ostream& os, SparseArrayClass* SArr ){ if ( SArr ) { int old_prec = os.precision(); os.precision(3); *************** *** 201,208 **** return os; }; ! class FeatureValue: public ValueClass ! { friend ostream& operator<<( ostream&, const FeatureValue * ); public: FeatureValue( HashInfo *, int ); --- 199,206 ---- return os; }; ! class FeatureValue: public ValueClass { ! friend class Feature; friend ostream& operator<<( ostream&, const FeatureValue * ); public: FeatureValue( HashInfo *, int ); *************** *** 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]; }; }; --- 224,242 ---- ValueClass **ValuesArray; BaseFeatTargClass( int, int, StringHash * ); ~BaseFeatTargClass(); ! int ArraySize() { return array_size; }; ! int EffectiveValues(); ! unsigned int TotalValues(); ValueClass *Lookup( char * ); protected: int CurSize; int Increment; ! int array_size; StringHash *TokenTree; IImaptype Mapping; void enlarge_values_array( int ); ! int localmapping( int ); ! ValueClass *get_value_index( int I ) { return ValuesArray[I]; }; }; *************** *** 247,253 **** class Target: public BaseFeatTargClass { public: Target( int a, int b, StringHash *T ): BaseFeatTargClass(a,b,T) {}; ! TargetValue *add_value( char * ); }; class Feature: public BaseFeatTargClass { --- 243,251 ---- class Target: public BaseFeatTargClass { public: Target( int a, int b, StringHash *T ): BaseFeatTargClass(a,b,T) {}; ! TargetValue *add_value( char *, int freq = 1 ); ! bool decrement_value( TargetValue * ); ! bool increment_value( TargetValue * ); }; class Feature: public BaseFeatTargClass { *************** *** 288,293 **** --- 286,293 ---- }; ~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 v3/MBLClass.cxx *** v0/MBLClass.cxx Tue Sep 5 11:22:36 2000 --- v3/MBLClass.cxx Wed Sep 20 16:44:39 2000 *************** *** 18,23 **** --- 18,31 ---- * 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 + // 20-09-2000 Ko : made LearningInfo global and correct (i hope). + // + #include #include #include *************** *** 415,421 **** return true; }; ! #endif bool MBLClass::ShowWeights( ostream &os ) { --- 423,429 ---- return true; }; ! #endif // BUILD_SERVER bool MBLClass::ShowWeights( ostream &os ) { *************** *** 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 ) --- 508,520 ---- 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 ) *************** *** 565,572 **** void MBLClass::InitWeights(void) { ! if ( TreeOrder == DataFile ) ! Weighting = UserDefined; int i; for ( i=0; i< NumOfFeatures; i++ ){ if ( Features[i]->Ignore ) --- 569,576 ---- void MBLClass::InitWeights(void) { ! // if ( TreeOrder == DataFile ) ! // Weighting = UserDefined; int i; for ( i=0; i< NumOfFeatures; i++ ){ if ( Features[i]->Ignore ) *************** *** 709,715 **** time_stamp( "Calculating Entropy " ); calculate_fv_entropy( false ); if ( verbosity & FEAT_W ){ ! learning_info( cerr, DataLines, SkippedLines ); result = true; } } --- 713,722 ---- time_stamp( "Calculating Entropy " ); calculate_fv_entropy( false ); if ( verbosity & FEAT_W ){ ! cerr << "Lines of data : " << DataLines << endl; ! if ( SkippedLines != 0 ) ! cerr << "SkippedLines : " << SkippedLines << endl; ! LearningInfo( cerr ); result = true; } } *************** *** 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; --- 750,768 ---- 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 ) --- 980,989 ---- // 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 ) { --- 1004,1010 ---- 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 = --- 1011,1017 ---- 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( ); --- 1041,1049 ---- 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; --- 1332,1338 ---- 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; *************** *** 1410,1425 **** } } ! void MBLClass::learning_info( ostream& s, int Total, int Skipped ) ! { int i; ! s << "Lines of data : " << Total << endl; ! if ( Skipped != 0 ) ! s << "SkippedLines : " << Skipped << endl; 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; --- 1417,1430 ---- } } ! void MBLClass::LearningInfo( ostream& s ){ int i; ! calculate_fv_entropy( true ); 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 --- 1438,1445 ---- 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( ); --- 1476,1482 ---- 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); --- 1845,1854 ---- 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. // --- 1968,1977 ---- // 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; } } --- 1997,2006 ---- // 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; } --- 2456,2471 ---- 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, --- 2475,2483 ---- 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 ){ --- 2486,2493 ---- 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; } --- 2500,2509 ---- ResultDist = test_instance_full( CurrInst, Distance ); } UnHideInstance( CurrInst ); + MBLInit = false; } else { ! FatalError( " Hiding failed!\n" ); } return ResultDist; } *************** *** 3075,3080 **** --- 3076,3083 ---- if ( isspace(Buffer[j]) ){ result++; while ( isspace( Buffer[++j] ) ); + if ( Buffer[j] == '\0' ) + result--; // we had some trailing spaces } }; break; diff -c v0/MBLClass.h v3/MBLClass.h *** v0/MBLClass.h Tue Sep 5 10:51:22 2000 --- v3/MBLClass.h Wed Sep 20 16:44:45 2000 *************** *** 123,128 **** --- 123,129 ---- void FatalError( const char*... ); int SockNum() const { return Socket; }; int MaxFeats() { return MaxFeatures; }; + void LearningInfo( ostream& ); protected: MBLClass() {}; MBLClass( AlgorithmType, const int N = 1000, const char * = NULL ); *************** *** 241,247 **** inline bool set_input_format( InputFormatType ); inline void show_input_format( ); inline void set_order(void); - void learning_info( ostream&, int, int ); void calc_perm( double * ); inline void write_perm(void); inline void write_perm_special(ostream&); --- 242,247 ---- diff -c v0/Makefile v3/Makefile *** v0/Makefile Tue Sep 5 11:22:37 2000 --- v3/Makefile Tue Sep 5 11:41:09 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) diff -c v0/Options.h v3/Options.h *** v0/Options.h Tue Sep 5 10:51:33 2000 --- v3/Options.h Wed Sep 20 16:44:45 2000 *************** *** 16,21 **** --- 16,23 ---- * Timbl@kub.nl */ + // Bug fixed 23-08-2000 Ko: Free-in the result of verbosity_to_string + #ifndef OPTIONS_H #define OPTIONS_H *************** *** 108,122 **** return result; }; ostream& show_opt( ostream &os ){ os.width(20); os.setf( ios::left, ios::adjustfield ); ! os << Name << " : " << verbosity_to_string(*V); return os; }; ostream& show_full( ostream &os ){ os.width(20); os.setf( ios::left, ios::adjustfield ); ! os << Name << " : " << verbosity_to_string(*V); return os; }; }; --- 110,128 ---- return result; }; ostream& show_opt( ostream &os ){ + const char *tmp = verbosity_to_string(*V); os.width(20); os.setf( ios::left, ios::adjustfield ); ! os << Name << " : " << tmp; ! free( const_cast(tmp) ); return os; }; ostream& show_full( ostream &os ){ + const char *tmp = verbosity_to_string(*V); os.width(20); os.setf( ios::left, ios::adjustfield ); ! os << Name << " : " << tmp; ! free( const_cast(tmp) ); return os; }; }; Common subdirectories: v0/SPARC and v3/SPARC diff -c v0/ServerProcs.cxx v3/ServerProcs.cxx *** v0/ServerProcs.cxx Tue Sep 5 10:51:34 2000 --- v3/ServerProcs.cxx Wed Sep 20 16:44:39 2000 *************** *** 19,24 **** --- 19,25 ---- * Timbl@kub.nl */ + #ifdef BUILD_SERVER #define _REENTRANT #include #include *************** *** 455,457 **** --- 456,459 ---- } } + #endif // BUILD_SERVER diff -c v0/SocketBasics.cxx v3/SocketBasics.cxx *** v0/SocketBasics.cxx Tue Sep 5 10:51:34 2000 --- v3/SocketBasics.cxx Wed Sep 20 16:44:39 2000 *************** *** 17,23 **** * or send mail to: * Timbl@kub.nl */ ! #include #include #include --- 17,23 ---- * or send mail to: * Timbl@kub.nl */ ! #ifdef BUILD_SERVER #include #include #include *************** *** 233,235 **** --- 233,236 ---- return write_line( socknum, line.c_str() ); } + #endif // BUILD_SERVER diff -c v0/Timbl.cxx v3/Timbl.cxx *** v0/Timbl.cxx Tue Sep 5 11:22:38 2000 --- v3/Timbl.cxx Wed Sep 20 16:44:39 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" *************** *** 91,99 **** cerr << "nothing to test or train!\n" << endl; exit(1); } ! if ( !GlobalOptPars.definitive_options( Run, false ) || !Run->GetInstanceBase( GlobalOptPars.TreeInFileName ) ) exit(1); if ( GlobalOptPars.NamesFileName != "" ) Run->WriteNamesFile( GlobalOptPars.NamesFileName ); bool more_to_do; --- 91,102 ---- cerr << "nothing to test or train!\n" << endl; exit(1); } ! if ( !GlobalOptPars.definitive_options( Run, true ) || !Run->GetInstanceBase( GlobalOptPars.TreeInFileName ) ) exit(1); + // Run->LearningInfo( cerr ); + if ( GlobalOptPars.SaveTree() ) + Run->WriteInstanceBase( GlobalOptPars.TreeOutFileName ); if ( GlobalOptPars.NamesFileName != "" ) Run->WriteNamesFile( GlobalOptPars.NamesFileName ); bool more_to_do; *************** *** 132,141 **** } while ( more_to_do ); } else { - char optline[3 * OPT_LENGTH]; - sprintf( optline, "TREE_ORDER: %s", - order_to_string(GlobalOptPars.Order()) ); - Run->SetOption( optline ); if ( !GlobalOptPars.definitive_options( Run, true ) ) exit(1); Run->PrepareExperiment( GlobalOptPars.DataFileName ); --- 135,140 ---- diff -c v0/TimblClient.cxx v3/TimblClient.cxx *** v0/TimblClient.cxx Tue Sep 5 10:51:35 2000 --- v3/TimblClient.cxx Wed Sep 20 16:44:39 2000 *************** *** 26,34 **** #include #include #ifdef BUILD_SERVER - #include "TimblExperiment.h" #include "ServerProcs.h" #else void RunClient( istream& Input, ostream& Output, --- 26,34 ---- #include #include + #include "TimblExperiment.h" #ifdef BUILD_SERVER #include "ServerProcs.h" #else void RunClient( istream& Input, ostream& Output, diff -c v0/Types.cxx v3/Types.cxx *** v0/Types.cxx Tue Sep 5 10:51:35 2000 --- v3/Types.cxx Wed Sep 20 16:44:39 2000 *************** *** 19,24 **** --- 19,26 ---- * Timbl@kub.nl */ + // bug fixed 23-08-2000 Ko: verbosity_to_string now returns a copy of c_str() + #include #include #include *************** *** 275,281 **** Outline += VerbStrings[i][(full?1:0)]; } } ! return Outline.c_str(); } const char *verbosity_to_string( VerbosityFlags v, bool full ) --- 277,283 ---- Outline += VerbStrings[i][(full?1:0)]; } } ! return strdup( Outline.c_str() ); } const char *verbosity_to_string( VerbosityFlags v, bool full ) Common subdirectories: v0/X86 and v3/X86 Only in v0: makefile Only in v3: makefile.ilk Only in v3: patch300-301 Only in v3: patch300-302 Only in v3: patch301-302