void TMultinomial::AddFtr(const TStrV& StrV, const TFltV& FltV, TIntFltKdV& SpV) const {
// make sure we either do not have explicit values, or their dimension matches with string keys
EAssertR(FltV.Empty() || (StrV.Len() == FltV.Len()), "TMultinomial::AddFtr:: String and double values not aligned");
// generate internal feature vector
SpV.Gen(StrV.Len(), 0);
for (int StrN = 0; StrN < StrV.Len(); StrN++) {
const int FtrId = FtrGen.GetFtr(StrV[StrN]);
// only use features we've seen during updates
if (FtrId != -1) {
const double Val = FltV.Empty() ? 1.0 : FltV[StrN].Val;
if (Val > 1e-16) { SpV.Add(TIntFltKd(FtrId, Val)); }
}
}
SpV.Sort();
// merge elements with the same id
int GoodSpN = 0;
for (int SpN = 1; SpN < SpV.Len(); SpN++) {
if (SpV[GoodSpN].Key == SpV[SpN].Key) {
// repetition of previous id, sum counts
SpV[GoodSpN].Dat += SpV[SpN].Dat;
} else {
// increase the pointer to the next good position
GoodSpN++;
// and move the new value down to the good position
SpV[GoodSpN] = SpV[SpN];
}
}
// truncate the vector
SpV.Trunc(GoodSpN + 1);
// replace values with 1 if needed
if (IsBinary()) { for (TIntFltKd& Sp : SpV) { Sp.Dat = 1.0; } }
// final normalization, if needed
if (IsNormalize()) { TLinAlg::Normalize(SpV); }
}
void TJsonVal::GetArrNumSpV(TIntFltKdV& NumSpV) const {
EAssert(IsArr());
for (int ElN = 0; ElN < GetArrVals(); ElN++) {
PJsonVal ArrVal = GetArrVal(ElN);
EAssert(ArrVal->IsArr());
EAssert(ArrVal->GetArrVals() == 2);
int Idx = ArrVal->GetArrVal(0)->GetInt();
double Val = ArrVal->GetArrVal(1)->GetNum();
NumSpV.Add(TIntFltKd(Idx, Val));
}
NumSpV.Sort();
}
void TStrFeatureSpace::FromAddStr(const TStr& Serialized, TIntFltKdV& Vec, char Sep) {
TStrV Toks;
Serialized.SplitOnAllCh(Sep, Toks, true);
Vec.Gen(Toks.Len());
for (int i = 0; i < Toks.Len(); i++) {
TStr Key, Value;
Toks[i].SplitOnCh(Key, ':', Value);
int FeatureId = GetAddId(Key);
double FeatureWgt;
if (Value.IsFlt(FeatureWgt)) {
Vec[i].Key = FeatureId;
Vec[i].Dat = FeatureWgt;
} else {
EFailR((Value + TStr(" is not a valid floating point number.")).CStr());
}
}
Vec.Sort();
}
void TStrFeatureSpace::FromStr(const TStr& Serialized, TIntFltKdV& Vec, char Sep) const {
TStrV Toks;
Serialized.SplitOnAllCh(Sep, Toks, true);
Vec.Gen(Toks.Len(),0);
for (int i = 0; i < Toks.Len(); i++) {
TStr Key, Value;
Toks[i].SplitOnCh(Key, ':', Value);
TStrFSSize FeatureId;
if (GetIfExistsId(Key, FeatureId)) {
double FeatureWgt;
if (Value.IsFlt(FeatureWgt)) {
TIntFltKd& Kv = Vec[Vec.Add()];
Kv.Key = FeatureId;
Kv.Dat = FeatureWgt;
} else {
EFailR((Value + TStr(" is not a valid floating point number.")).CStr());
}
}
}
Vec.Sort();
}
void TBagOfWords::AddFtr(const TStrV& TokenStrV, TIntFltKdV& SpV) const {
// aggregate token counts
TIntH TermFqH;
TStrV NgramStrV;
GenerateNgrams(TokenStrV, NgramStrV);
for (int TokenStrN = 0; TokenStrN < NgramStrV.Len(); TokenStrN++) {
const TStr& TokenStr = NgramStrV[TokenStrN];
// get token ID
const int TokenId = IsHashing() ?
(TokenStr.GetHashTrick() % HashDim) : // hashing
TokenSet.GetKeyId(TokenStr); // vocabulary
// add if known token
if (TokenId != -1) {
TermFqH.AddDat(TokenId)++;
}
}
// make a sparse vector out of it
SpV.Gen(TermFqH.Len(), 0);
int KeyId = TermFqH.FFirstKeyId();
while (TermFqH.FNextKeyId(KeyId)) {
const int TermId = TermFqH.GetKey(KeyId);
double TermVal = 1.0;
if (IsTf()) { TermVal *= double(TermFqH[KeyId]); }
if (IsIdf()) {
if (ForgetP) {
const double DocFq = double(DocFqV[TermId]) + OldDocFqV[TermId];
if (DocFq > 0.1) { TermVal *= log((double(Docs) + OldDocs) / DocFq); }
} else {
TermVal *= log(double(Docs) / double(DocFqV[TermId]));
}
}
SpV.Add(TIntFltKd(TermId, TermVal));
}
SpV.Sort();
// step (4): normalize the vector if so required
if (IsNormalize()) { TLinAlg::Normalize(SpV); }
}
