void TSkyGridEnt::GetSorted_LinkWgtDstEntIdPrV(
const uint64& MnTm, const double& TopWgtSumPrc, TIntPrV& LinkWgtDstEntIdPrV) const {
double AllLinkWgtSum=0;
TIntIntH DstEntIdLinkWgtH;
int LinkEnts=GetLinkEnts();
for (int LinkEntN=0; LinkEntN<LinkEnts; LinkEntN++){
int DstEntId=GetLinkEntId(LinkEntN);
int EntLinks=GetEntLinks(LinkEntN);
int EntLinkWgtSum=0;
for (int EntLinkN=0; EntLinkN<EntLinks; EntLinkN++){
const TSkyGridEntLinkCtx& EntLinkCtx=GetEntLinkCtx(LinkEntN, EntLinkN);
if (EntLinkCtx.Tm>=MnTm){
EntLinkWgtSum+=EntLinkCtx.LinkWgt;}
}
DstEntIdLinkWgtH.AddDat(DstEntId, EntLinkWgtSum);
AllLinkWgtSum+=EntLinkWgtSum;
}
LinkWgtDstEntIdPrV.Clr(); DstEntIdLinkWgtH.GetDatKeyPrV(LinkWgtDstEntIdPrV);
LinkWgtDstEntIdPrV.Sort(false);
// cut long-tail
if ((TopWgtSumPrc>0.0)&&(LinkWgtDstEntIdPrV.Len()>0)){
int TopLinkWgt=LinkWgtDstEntIdPrV[0].Val1;
if (TopLinkWgt>(3*AllLinkWgtSum)/LinkWgtDstEntIdPrV.Len()){
double CutWgtSum=AllLinkWgtSum*(1-TopWgtSumPrc);
int LastValN=LinkWgtDstEntIdPrV.Len()-1;
while ((LastValN>0)&&(CutWgtSum>0)){
CutWgtSum-=LinkWgtDstEntIdPrV[LastValN].Val1;
LastValN--;
}
LinkWgtDstEntIdPrV.Trunc(LastValN+1);
}
}
}
void TSkyGridBs::GetLinkWgtDstEntIdPrVDiff(
const TIntPrV& OldLinkWgtDstEntIdPrV, const TIntPrV& NewLinkWgtDstEntIdPrV,
TIntPrV& NegDiffLinkWgtDstEntIdPrV, TIntPrV& PosDiffLinkWgtDstEntIdPrV){
TIntIntH DstEntIdToLinkWgtH;
// set previous-vector
for (int WordN=0; WordN<NewLinkWgtDstEntIdPrV.Len(); WordN++){
int LinkWgt=NewLinkWgtDstEntIdPrV[WordN].Val1;
int DstEntId=NewLinkWgtDstEntIdPrV[WordN].Val2;
DstEntIdToLinkWgtH.AddDat(DstEntId, LinkWgt);
}
// diff current-vector
for (int WordN=0; WordN<OldLinkWgtDstEntIdPrV.Len(); WordN++){
int LinkWgt=OldLinkWgtDstEntIdPrV[WordN].Val1;
int DstEntId=OldLinkWgtDstEntIdPrV[WordN].Val2;
int CurLinkWgt=DstEntIdToLinkWgtH.AddDat(DstEntId);
DstEntIdToLinkWgtH.AddDat(DstEntId, CurLinkWgt-LinkWgt);
}
// extract vector
TIntPrV _DiffLinkWgtDstEntIdPrV;
DstEntIdToLinkWgtH.GetDatKeyPrV(_DiffLinkWgtDstEntIdPrV);
// clean zeros
TIntPrV DiffLinkWgtDstEntIdPrV(_DiffLinkWgtDstEntIdPrV.Len(), 0);
for (int EntN=0; EntN<_DiffLinkWgtDstEntIdPrV.Len(); EntN++){
int LinkWgt=_DiffLinkWgtDstEntIdPrV[EntN].Val1;
if (LinkWgt!=0){
DiffLinkWgtDstEntIdPrV.Add(_DiffLinkWgtDstEntIdPrV[EntN]);}
}
// positive-vector
DiffLinkWgtDstEntIdPrV.Sort(true);
NegDiffLinkWgtDstEntIdPrV=DiffLinkWgtDstEntIdPrV;
// negative-vector
DiffLinkWgtDstEntIdPrV.Sort(false);
PosDiffLinkWgtDstEntIdPrV=DiffLinkWgtDstEntIdPrV;
}
// Test the default constructor
TEST(TIntIntH, DefaultConstructor) {
TIntIntH TableInt;
EXPECT_EQ(1,TableInt.Empty());
EXPECT_EQ(0,TableInt.Len());
EXPECT_EQ(0,TableInt.GetMxKeyIds());
}
void TNmObjBs::GetFqNmObjIdPrV(
const TStr& TargetNmObjStr, TIntPrV& FqNmObjIdPrV) const {
//printf("Searching %s ...", TargetNmObjStr.CStr());
// get target named-object-id
int TargetNmObjId=GetNmObjId(TargetNmObjStr);
// collect target named-object frequencies
TIntIntH NmObjIdToFqH;
// traverse target named-object documents
int NmObjDocs=GetNmObjDocs(TargetNmObjId);
for (int NmObjDocIdN=0; NmObjDocIdN<NmObjDocs; NmObjDocIdN++){
// get document-id
int DocId=GetNmObjDocId(TargetNmObjId, NmObjDocIdN);
// traverse named-object in document
int DocNmObjs=GetDocNmObjs(DocId);
for (int DocNmObjN=0; DocNmObjN<DocNmObjs; DocNmObjN++){
// get named-object & frequency
int NmObjId; int TermFq;
GetDocNmObjId(DocId, DocNmObjN, NmObjId, TermFq);
// increment named-object document frequency
NmObjIdToFqH.AddDat(NmObjId)++;
}
}
// get & sort frequency table
FqNmObjIdPrV.Clr(); NmObjIdToFqH.GetDatKeyPrV(FqNmObjIdPrV);
FqNmObjIdPrV.Sort(false);
}
/////////////////////////////////////////////////
// SkyGrid-Base
void TSkyGridBs::GetSorted_DocsEntIdPrV(TIntPrV& DocsEntIdPrV){
TIntIntH EntIdToDocsH;
for (int EntId=0; EntId<GetEnts(); EntId++){
int Docs=GetEnt(EntId).GetDocIds();
EntIdToDocsH.AddDat(EntId, Docs);
}
DocsEntIdPrV.Clr(); EntIdToDocsH.GetDatKeyPrV(DocsEntIdPrV);
DocsEntIdPrV.Sort(false);
}
TIntH LoadNodeList(TStr InFNmNodes) {
TSsParser Ss(InFNmNodes, ssfWhiteSep, true, true, true);
TIntIntH Nodes;
int br = 0, NId;
while (Ss.Next()) {
if (Ss.GetInt(0, NId)) {
Nodes.AddDat(br, NId);
br++;
}
}
return Nodes;
}
void TCycBs::_SaveTaxonomyTxt(FILE* fOut,
const int& Lev, TIntPrV& RelIdVIdPrV, TIntIntH& VIdToLevH){
for (int VidN=0; VidN<RelIdVIdPrV.Len(); VidN++){
int FromRelId=RelIdVIdPrV[VidN].Val1;
int SrcVId=RelIdVIdPrV[VidN].Val2;
TStr SrcVNm=GetVNm(SrcVId);
TCycVrtx& SrcVrtx=GetVrtx(SrcVId);
if (!SrcVrtx.IsFlag(cvfHumanOk)){continue;}
TStr FlagStr=SrcVrtx.GetFlagStr();
if (FromRelId==-1){
if (Lev>0){fprintf(fOut, "===upper");} else {fprintf(fOut, "===lower");}
fprintf(fOut, "=======================================================\n");
fprintf(fOut, "%s - %s\n", SrcVNm.CStr(), FlagStr.CStr());
} else {
TStr FromRelNm=GetVNm(FromRelId);
fprintf(fOut, "%*c[%s] --> %s\n", (Lev-1)*5, ' ', FromRelNm.CStr(), SrcVNm.CStr());
}
TIntPrV UpRelIdVIdPrV;
for (int EdgeN=0; EdgeN<SrcVrtx.GetEdges(); EdgeN++){
TCycEdge& Edge=SrcVrtx.GetEdge(EdgeN);
int RelId=Edge.GetRelId();
int DstVId=Edge.GetDstVId();
TStr RelNm=GetVNm(RelId);
TStr DstVNm=GetVNm(DstVId);
if (Lev>0){
// upper taxonomy
if ((RelNm=="#$isa")||(RelNm=="#$genls")){
if (!VIdToLevH.IsKey(DstVId)){
VIdToLevH.AddDat(DstVId, Lev+1);
UpRelIdVIdPrV.Add(TIntPr(RelId, DstVId));
}
}
} else {
// lower taxonomy
if ((RelNm=="~#$isa")||(RelNm=="~#$genls")){
if (!VIdToLevH.IsKey(DstVId)){
VIdToLevH.AddDat(DstVId, Lev-1);
UpRelIdVIdPrV.Add(TIntPr(RelId, DstVId));
}
}
}
}
// recursive call
if (Lev>0){
_SaveTaxonomyTxt(fOut, Lev+1, UpRelIdVIdPrV, VIdToLevH);
} else {
_SaveTaxonomyTxt(fOut, Lev-1, UpRelIdVIdPrV, VIdToLevH);
}
}
}
void TCycBs::SaveTaxonomyTxt(const TStr& FNm){
TFOut FOut(FNm); FILE* fOut=FOut.GetFileId();
for (int VId=0; VId<GetVIds(); VId++){
printf("%d/%d (%.1f%%)\r", 1+VId, GetVIds(), 100.0*(1+VId)/GetVIds());
//if (VId>10){break;}
// upper taxonomy
{int Lev=0;
TIntIntH VIdToLevH; VIdToLevH.AddDat(VId, Lev);
TIntPrV UpRelIdVIdPrV; UpRelIdVIdPrV.Add(TIntPr(-1, VId));
_SaveTaxonomyTxt(fOut, Lev+1, UpRelIdVIdPrV, VIdToLevH);}
// lower taxonomy
{int Lev=0;
TIntIntH VIdToLevH; VIdToLevH.AddDat(VId, Lev);
TIntPrV UpRelIdVIdPrV; UpRelIdVIdPrV.Add(TIntPr(-1, VId));
_SaveTaxonomyTxt(fOut, Lev-1, UpRelIdVIdPrV, VIdToLevH);}
}
printf("\n");
}
double BorgattiEverettMeasure(PUNGraph& Graph, TIntIntH& out, double coresize, int type) {
double sum = 0.0;
for (TUNGraph::TEdgeI EI = Graph->BegEI(); EI < Graph->EndEI(); EI++) { // Calculate and store the degrees of each node.
int i = EI.GetSrcNId();
int j = EI.GetDstNId();
if (type == 1) {
if (out.GetDat(i) == 1 || out.GetDat(j) == 1)
sum += 1;
}
else {
if (out.GetDat(i) == 1 && out.GetDat(j) == 1)
sum += 1;
}
}
return sum/(((coresize*coresize)-coresize)/2);
}
double PearsonCorrelation(PUNGraph& Graph, TIntIntH& out, int coresize) {
int br_core1=0,br_periphery1=0,br_core_per1=0;
for (TUNGraph::TEdgeI EI = Graph->BegEI(); EI < Graph->EndEI(); EI++) { // Calculate and store the degrees of each node.
int i = EI.GetSrcNId();
int j = EI.GetDstNId();
if (out.GetDat(i)==1&&out.GetDat(j)==1 && i!=j)
br_core1++;
else if (out.GetDat(i)==0&&out.GetDat(j)==0 && i!=j)
br_periphery1++;
else
br_core_per1++;
}
double core_quality = (double)br_core1/((((double)coresize*(double)coresize)-(double)coresize)/2);
int per_size = Graph->GetNodes()-coresize;
double periphery_quality = (((((double)per_size*(double)per_size)-(double)per_size)/2) - (double)br_periphery1)/((((double)per_size*(double)per_size)-(double)per_size)/2);
return (double)(core_quality+periphery_quality);
}
// hash table benchmark with integer keys
void HashBench(const int& n) {
TIntIntH TableInt;
float ft0, ft1;
int x;
int i;
int Found;
int NotFound;
int Id;
// build the hash table
ft0 = GetCPUTime();
for (i = 0; i < n; i++) {
x = (int) (drand48() * 100000000);
TableInt.AddDat(x,0);
}
printf("hash: size %d\n", TableInt.Len());
ft1 = GetCPUTime();
printf("hash: %7.3fs inserting %d numbers\n",ft1-ft0,i);
// search the hash table
ft0 = GetCPUTime();
Found = 0;
NotFound = 0;
for (i = 0; i < n; i++) {
x = (int) (drand48() * 100000000);
Id = TableInt.GetKeyId(x);
if (Id < 0) {
NotFound++;
} else {
Found++;
}
}
printf("hash: found %d, notfound %d\n", Found, NotFound);
ft1 = GetCPUTime();
printf("hash: %7.3fs searching %d numbers\n",ft1-ft0,i);
}
void LearnEmbeddings(TVVec<TInt, int64>& WalksVV, int& Dimensions, int& WinSize,
int& Iter, bool& Verbose, TIntFltVH& EmbeddingsHV) {
TIntIntH RnmH;
TIntIntH RnmBackH;
int64 NNodes = 0;
//renaming nodes into consecutive numbers
for (int i = 0; i < WalksVV.GetXDim(); i++) {
for (int64 j = 0; j < WalksVV.GetYDim(); j++) {
if ( RnmH.IsKey(WalksVV(i, j)) ) {
WalksVV(i, j) = RnmH.GetDat(WalksVV(i, j));
} else {
RnmH.AddDat(WalksVV(i,j),NNodes);
RnmBackH.AddDat(NNodes,WalksVV(i, j));
WalksVV(i, j) = NNodes++;
}
}
}
TIntV Vocab(NNodes);
LearnVocab(WalksVV, Vocab);
TIntV KTable(NNodes);
TFltV UTable(NNodes);
TVVec<TFlt, int64> SynNeg;
TVVec<TFlt, int64> SynPos;
TRnd Rnd(time(NULL));
InitPosEmb(Vocab, Dimensions, Rnd, SynPos);
InitNegEmb(Vocab, Dimensions, SynNeg);
InitUnigramTable(Vocab, KTable, UTable);
TFltV ExpTable(TableSize);
double Alpha = StartAlpha; //learning rate
#pragma omp parallel for schedule(dynamic)
for (int i = 0; i < TableSize; i++ ) {
double Value = -MaxExp + static_cast<double>(i) / static_cast<double>(ExpTablePrecision);
ExpTable[i] = TMath::Power(TMath::E, Value);
}
int64 WordCntAll = 0;
// op RS 2016/09/26, collapse does not compile on Mac OS X
//#pragma omp parallel for schedule(dynamic) collapse(2)
for (int j = 0; j < Iter; j++) {
#pragma omp parallel for schedule(dynamic)
for (int64 i = 0; i < WalksVV.GetXDim(); i++) {
TrainModel(WalksVV, Dimensions, WinSize, Iter, Verbose, KTable, UTable,
WordCntAll, ExpTable, Alpha, i, Rnd, SynNeg, SynPos);
}
}
if (Verbose) { printf("\n"); fflush(stdout); }
for (int64 i = 0; i < SynPos.GetXDim(); i++) {
TFltV CurrV(SynPos.GetYDim());
for (int j = 0; j < SynPos.GetYDim(); j++) { CurrV[j] = SynPos(i, j); }
EmbeddingsHV.AddDat(RnmBackH.GetDat(i), CurrV);
}
}
int FastCorePeriphery(PUNGraph& Graph, TIntIntH& out) {
TIntIntH nodes;
double Z=0;
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) { // Calculate and store the degrees of each node.
int deg = NI.GetDeg();
int id = NI.GetId();
Z += deg;
nodes.AddDat(id,deg);
}
Z = Z/2;
nodes.SortByDat(false); // Then sort the nodes in descending order of degree, to get a list of nodes {v1, v2, . . . , vn}.
double Zbest = 99999900000000000;
int kbest = 0;
int br=0;
for (int k=0; k<nodes.Len(); k++) {
br++;
Z = Z + br - 1 - nodes[k];
if (Z < Zbest) { // or <=
Zbest = Z;
kbest = br;
}
}
int cp = 0;
br = 0;
for (THashKeyDatI<TInt, TInt> it = nodes.BegI(); !it.IsEnd(); it++) {
if (br < kbest)
cp = 1;
else
cp = 0;
out.AddDat(it.GetKey(), cp);
br++;
}
return kbest;
}
void TGreedyAlg::generateCascades(const int& noCasacdes, const double& pInit, const double& p, const double& q) {
int noNodes = groundTruthGraph->GetNodes();
// printf("Generating cascade for graph with noNodes = %d:\n\n", noNodes);
if (noNodes == 0) {
return;
}
// set random seed
TInt::Rnd.Randomize();
for (int casacdeI = 0; casacdeI < noCasacdes; casacdeI++) {
TCascade cascade;
double globalTime = 0;
int noActiveNodes = 0;
int noSusceptibleNodes = 0;
int noInactiveNodes = 0;
TIntIntH nodeStates;
// flip biased coin for each node to collect seeds
for (TKColourNet::TNodeI NI = groundTruthGraph->BegNI(); NI < groundTruthGraph->EndNI(); NI++) {
const int nodeId = NI.GetId();
double flipResult = TInt::Rnd.GetUniDev();
// printf("nodeId = %d, flipResult = %f\n", nodeId, (float) flipResult);
if (flipResult <= pInit) {
nodeStates.AddDat(nodeId) = activeState;
// printf("##### ADD TO CASCADE: nodeId = %d, globalTime = %f #####\n", nodeId, globalTime);
cascade.Add(nodeId, globalTime);
noActiveNodes++;
}
else {
nodeStates.AddDat(nodeId) = susceptibleState;
noSusceptibleNodes++;
}
}
globalTime++;
while (noActiveNodes > 0) {
// printf("\n***** noActiveNodes = %d, noSusceptibleNodes = %d, noInactiveNodes = %d, globalTime = %f *****\n\n", noActiveNodes, noSusceptibleNodes, noInactiveNodes, (float) globalTime);
const TIntIntH beginningNodeStates = nodeStates;
// for each node in the graph
for (TKColourNet::TNodeI NI = groundTruthGraph->BegNI(); NI < groundTruthGraph->EndNI(); NI++) {
const int nodeId = NI.GetId();
const int nodeColourId = NI.GetDat().getColourId();
const int nodeState = beginningNodeStates.GetDat(nodeId);
// printf("nodeId = %d, nodeColourId = %d, nodeState = %d\n", nodeId, nodeColourId, nodeState);
// if node is active, infect susceptible child nodes with probability pij
if (nodeState == activeState) {
const TKColourNet::TNodeI LNI = groundTruthGraph->GetNI(nodeId);
for (int e = 0; e < LNI.GetOutDeg(); e++) {
const int childNodeId = LNI.GetOutNId(e);
const int childNodeColourId = LNI.GetOutNDat(e).getColourId();
const int childNodeState = nodeStates.GetDat(childNodeId);
// printf("childNodeId = %d, childNodeColourId = %d, childNodeState = %d\n", childNodeId, childNodeColourId, childNodeState);
if (childNodeState == susceptibleState) {
double probablityInfection = (nodeColourId == childNodeColourId) ? p : q;
double flipResult = TInt::Rnd.GetUniDev();
// printf("childNodeId = %d, probabilityInfection = %f, flipResult = %f\n", childNodeId, (float) probablityInfection, (float) flipResult);
// infection occurred: childNode goes from susceptible -> active
if (flipResult <= probablityInfection) {
nodeStates[childNodeId] = activeState;
// printf("##### ADD TO CASCADE: nodeId = %d, globalTime = %f #####\n", childNodeId, globalTime);
cascade.Add(childNodeId, globalTime);
noActiveNodes++;
noSusceptibleNodes--;
}
}
}
// main node goes from active -> inactive
nodeStates[nodeId] = inactiveState;
noActiveNodes--;
noInactiveNodes++;
}
}
globalTime++;
}
addCascade(cascade);
}
}
// Table manipulations
TEST(TIntIntH, ManipulateTable) {
const int64 NElems = 1000000;
int DDist = 10;
const char *FName = "test.hashint.dat";
TIntIntH TableInt;
TIntIntH TableInt1;
TIntIntH TableInt2;
int i;
int d;
int n;
int Id;
int Key;
int64 KeySumVal;
int64 DatSumVal;
int64 KeySum;
int64 DatSum;
int64 KeySumDel;
int64 DatSumDel;
int DelCount;
int Count;
// add table elements
d = Prime(NElems);
n = d;
KeySumVal = 0;
DatSumVal = 0;
for (i = 0; i < NElems; i++) {
TableInt.AddDat(n,n+1);
KeySumVal += n;
DatSumVal += (n+1);
//printf("add %d %d\n", n, n+1);
n = (n + d) % NElems;
}
EXPECT_EQ(0,TableInt.Empty());
EXPECT_EQ(NElems,TableInt.Len());
EXPECT_EQ(0,(NElems-1)*(NElems)/2 - KeySumVal);
EXPECT_EQ(0,(NElems)*(NElems+1)/2 - DatSumVal);
// verify elements by successive keys
KeySum = 0;
DatSum = 0;
for (i = 0; i < NElems; i++) {
Id = TableInt.GetKeyId(i);
EXPECT_EQ(1,Id >= 0);
Key = TableInt.GetKey(Id);
EXPECT_EQ(0,TableInt.GetDat(Key)-Key-1);
KeySum += Key;
DatSum += TableInt.GetDat(Key);
}
EXPECT_EQ(0,KeySumVal - KeySum);
EXPECT_EQ(0,DatSumVal - DatSum);
// verify elements by distant keys
KeySum = 0;
DatSum = 0;
n = Prime(d);
for (i = 0; i < NElems; i++) {
Id = TableInt.GetKeyId(n);
EXPECT_EQ(1,Id >= 0);
Key = TableInt.GetKey(Id);
EXPECT_EQ(0,TableInt.GetDat(Key)-Key-1);
KeySum += Key;
DatSum += TableInt.GetDat(Key);
n = (n + d) % NElems;
}
EXPECT_EQ(0,KeySumVal - KeySum);
EXPECT_EQ(0,DatSumVal - DatSum);
// verify elements by iterator
KeySum = 0;
DatSum = 0;
for (TIntIntH::TIter It = TableInt.BegI(); It < TableInt.EndI(); It++) {
EXPECT_EQ(0,It.GetDat()-It.GetKey()-1);
KeySum += It.GetKey();
DatSum += It.GetDat();
}
EXPECT_EQ(0,KeySumVal - KeySum);
EXPECT_EQ(0,DatSumVal - DatSum);
// verify elements by key index
KeySum = 0;
DatSum = 0;
Id = TableInt.FFirstKeyId();
while (TableInt.FNextKeyId(Id)) {
EXPECT_EQ(1,Id >= 0);
Key = TableInt.GetKey(Id);
EXPECT_EQ(0,TableInt.GetDat(Key)-Key-1);
KeySum += Key;
DatSum += TableInt.GetDat(Key);
}
EXPECT_EQ(0,KeySumVal - KeySum);
EXPECT_EQ(0,DatSumVal - DatSum);
// delete elements
DelCount = 0;
KeySumDel = 0;
DatSumDel = 0;
for (n = 0; n < NElems; n += DDist) {
Id = TableInt.GetKeyId(n);
//printf("del %d %d %d\n", n, Id, (int) TableInt[Id]);
KeySumDel += n;
DatSumDel += TableInt[Id];
TableInt.DelKeyId(Id);
DelCount++;
}
EXPECT_EQ(0,TableInt.Empty());
EXPECT_EQ(NElems-DelCount,TableInt.Len());
// verify elements by iterator
KeySum = 0;
DatSum = 0;
Count = 0;
for (TIntIntH::TIter It = TableInt.BegI(); It < TableInt.EndI(); It++) {
EXPECT_EQ(0,It.GetDat()-It.GetKey()-1);
//printf("get %d %d\n", (int) It.GetKey(), (int) It.GetDat());
KeySum += It.GetKey();
DatSum += It.GetDat();
Count++;
}
EXPECT_EQ(NElems-DelCount,Count);
EXPECT_EQ(0,KeySumVal - KeySumDel - KeySum);
EXPECT_EQ(0,DatSumVal - DatSumDel - DatSum);
// assignment
TableInt1 = TableInt;
EXPECT_EQ(0,TableInt1.Empty());
EXPECT_EQ(NElems-DelCount,TableInt1.Len());
// verify elements by iterator
KeySum = 0;
DatSum = 0;
Count = 0;
for (TIntIntH::TIter It = TableInt1.BegI(); It < TableInt1.EndI(); It++) {
EXPECT_EQ(0,It.GetDat()-It.GetKey()-1);
//printf("get %d %d\n", (int) It.GetKey(), (int) It.GetDat());
KeySum += It.GetKey();
DatSum += It.GetDat();
Count++;
}
EXPECT_EQ(NElems-DelCount,Count);
EXPECT_EQ(0,KeySumVal - KeySumDel - KeySum);
EXPECT_EQ(0,DatSumVal - DatSumDel - DatSum);
// saving and loading
{
TFOut FOut(FName);
TableInt.Save(FOut);
FOut.Flush();
}
{
TFIn FIn(FName);
TableInt2.Load(FIn);
}
EXPECT_EQ(NElems-DelCount,TableInt2.Len());
// verify elements by iterator
KeySum = 0;
DatSum = 0;
Count = 0;
for (TIntIntH::TIter It = TableInt2.BegI(); It < TableInt2.EndI(); It++) {
EXPECT_EQ(0,It.GetDat()-It.GetKey()-1);
//printf("get %d %d\n", (int) It.GetKey(), (int) It.GetDat());
KeySum += It.GetKey();
DatSum += It.GetDat();
Count++;
}
EXPECT_EQ(NElems-DelCount,Count);
EXPECT_EQ(0,KeySumVal - KeySumDel - KeySum);
EXPECT_EQ(0,DatSumVal - DatSumDel - DatSum);
// remove all elements
for (i = 0; i < Count; i++) {
Id = TableInt.GetRndKeyId(TInt::Rnd, 0.5);
TableInt.DelKeyId(Id);
}
EXPECT_EQ(0,TableInt.Len());
EXPECT_EQ(1,TableInt.Empty());
// verify elements by iterator
KeySum = 0;
DatSum = 0;
Count = 0;
for (TIntIntH::TIter It = TableInt.BegI(); It < TableInt.EndI(); It++) {
EXPECT_EQ(0,It.GetDat()-It.GetKey()-1);
//printf("get %d %d\n", (int) It.GetKey(), (int) It.GetDat());
KeySum += It.GetKey();
DatSum += It.GetDat();
Count++;
}
EXPECT_EQ(0,Count);
EXPECT_EQ(0,KeySum);
EXPECT_EQ(0,DatSum);
// clear the table
TableInt1.Clr();
EXPECT_EQ(0,TableInt1.Len());
EXPECT_EQ(1,TableInt1.Empty());
// verify elements by iterator
KeySum = 0;
DatSum = 0;
Count = 0;
for (TIntIntH::TIter It = TableInt1.BegI(); It < TableInt1.EndI(); It++) {
EXPECT_EQ(0,It.GetDat()-It.GetKey()-1);
//printf("get %d %d\n", (int) It.GetKey(), (int) It.GetDat());
KeySum += It.GetKey();
DatSum += It.GetDat();
Count++;
}
EXPECT_EQ(0,Count);
EXPECT_EQ(0,KeySum);
EXPECT_EQ(0,DatSum);
}
PBowDocBs TBowFl::LoadSvmLightTxt(
const TStr& DocDefFNm, const TStr& WordDefFNm,
const TStr& TrainDataFNm, const TStr& TestDataFNm,
const int& MxDocs){ //TODO: use MxDocs
// prepare document set
PBowDocBs BowDocBs=TBowDocBs::New();
int MOneCId=BowDocBs->CatNmToFqH.AddKey("-1");
int POneCId=BowDocBs->CatNmToFqH.AddKey("+1");
// document definition
bool DocDefP=false;
if (!DocDefFNm.Empty()&&(TFile::Exists(DocDefFNm))){
// (DId "DoxNm"<eoln>)*
PSIn SIn=TFIn::New(DocDefFNm);
TILx Lx(SIn, TFSet()|iloRetEoln|iloSigNum|iloExcept);
Lx.GetSym(syInt, syEof);
while (Lx.Sym==syInt){
int DId=Lx.Int;
Lx.GetSym(syColon);
Lx.GetSym(syQStr); TStr DocNm=Lx.Str;
Lx.GetSym(syEoln);
Lx.GetSym(syInt, syEof);
int NewDId=BowDocBs->DocNmToDescStrH.AddKey(DocNm);
EAssertR(DId==NewDId, "Document-Ids don't match.");
}
DocDefP=true;
}
// word definition
if (!WordDefFNm.Empty()&&(TFile::Exists(WordDefFNm))){
BowDocBs->WordStrToDescH.AddDat("Undef").Fq=0; // ... to have WId==0
PSIn SIn=TFIn::New(WordDefFNm);
TILx Lx(SIn, TFSet()|iloRetEoln|iloSigNum|iloExcept);
Lx.GetSym(syQStr, syEof);
while (Lx.Sym==syQStr){
TStr WordStr=Lx.Str;
Lx.GetSym(syInt); int WId=Lx.Int;
Lx.GetSym(syInt); int WordFq=Lx.Int;
Lx.GetSym(syEoln);
Lx.GetSym(syQStr, syEof);
int NewWId=BowDocBs->WordStrToDescH.AddKey(WordStr);
EAssertR(WId==NewWId, "Word-Ids don't match.");
BowDocBs->WordStrToDescH[WId].Fq=WordFq;
}
}
// train & test data
int MxWId=-1; TIntIntH WIdToFqH;
// train data
if (!TrainDataFNm.Empty()){
PSIn SIn=TFIn::New(TrainDataFNm);
TILx Lx(SIn, TFSet()|iloCmtAlw|iloRetEoln|iloSigNum|iloExcept);
// skip comment lines
while (Lx.GetSym(syInt, syEoln, syEof)==syEoln){}
// parse data lines
while (Lx.Sym==syInt){
// document
TStr DocNm=TInt::GetStr(BowDocBs->GetDocs());
int DId;
if (DocDefP){
DId=BowDocBs->DocNmToDescStrH.GetKeyId(DocNm);
} else {
DId=BowDocBs->DocNmToDescStrH.AddKey(DocNm);
}
BowDocBs->TrainDIdV.Add(DId);
// category (class value)
int CId=(Lx.Int==-1) ? MOneCId : POneCId;
BowDocBs->DocCIdVV.Add(); IAssert(DId==BowDocBs->DocCIdVV.Len()-1);
BowDocBs->DocCIdVV.Last().Gen(1, 0);
BowDocBs->DocCIdVV.Last().Add(CId);
// words (attributes)
PBowSpV SpV=TBowSpV::New(DId);
BowDocBs->DocSpVV.Add(SpV); IAssert(DId==BowDocBs->DocSpVV.Len()-1);
Lx.GetSym(syInt, syEoln);
while (Lx.Sym==syInt){
int WId=Lx.Int;
Lx.GetSym(syColon);
Lx.GetSym(syFlt); double WordFq=Lx.Flt;
Lx.GetSym(syInt, syEoln);
SpV->AddWIdWgt(WId, WordFq);
if (MxWId==-1){MxWId=WId;} else {MxWId=TInt::GetMx(MxWId, WId);}
WIdToFqH.AddDat(WId)++;
}
if (!Lx.CmtStr.Empty()){
// change document name to 'N' if comment 'docDesc=N'
TStr CmtStr=Lx.CmtStr;
static TStr DocNmPrefixStr="docDesc=";
if (CmtStr.IsPrefix(DocNmPrefixStr)){
TStr NewDocNm=
TStr("D")+CmtStr.GetSubStr(DocNmPrefixStr.Len(), CmtStr.Len()-1);
BowDocBs->DocNmToDescStrH.DelKey(DocNm);
int NewDId=BowDocBs->DocNmToDescStrH.AddKey(NewDocNm);
IAssert(DId==NewDId);
}
}
SpV->Trunc();
while (Lx.GetSym(syInt, syEoln, syEof)==syEoln){}
}
}
// test data
if (!TestDataFNm.Empty()){
PSIn SIn=TFIn::New(TestDataFNm);
TILx Lx(SIn, TFSet()|iloCmtAlw|iloRetEoln|iloSigNum|iloExcept);
while (Lx.GetSym(syInt, syEoln, syEof)==syEoln){}
while (Lx.Sym==syInt){
// document
TStr DocNm=TInt::GetStr(BowDocBs->GetDocs());
int DId;
if (DocDefP){
DId=BowDocBs->DocNmToDescStrH.GetKeyId(DocNm);
} else {
DId=BowDocBs->DocNmToDescStrH.AddKey(DocNm);
}
BowDocBs->TestDIdV.Add(DId);
// category (class value)
int CId=(Lx.Int==-1) ? MOneCId : POneCId;
BowDocBs->DocCIdVV.Add(); IAssert(DId==BowDocBs->DocCIdVV.Len()-1);
BowDocBs->DocCIdVV.Last().Gen(1, 0);
BowDocBs->DocCIdVV.Last().Add(CId);
// words (attributes)
PBowSpV SpV=TBowSpV::New(DId);
BowDocBs->DocSpVV.Add(SpV); IAssert(DId==BowDocBs->DocSpVV.Len()-1);
Lx.GetSym(syInt, syEoln);
while (Lx.Sym==syInt){
int WId=Lx.Int;
Lx.GetSym(syColon);
Lx.GetSym(syFlt); double WordFq=Lx.Flt;
Lx.GetSym(syInt, syEoln);
SpV->AddWIdWgt(WId, WordFq);
if (MxWId==-1){MxWId=WId;} else {MxWId=TInt::GetMx(MxWId, WId);}
WIdToFqH.AddDat(WId)++;
}
if (!Lx.CmtStr.Empty()){
// change document name to 'N' if comment 'docDesc=N'
TStr CmtStr=Lx.CmtStr;
static TStr DocNmPrefixStr="docDesc=";
if (CmtStr.IsPrefix(DocNmPrefixStr)){
TStr NewDocNm=
TStr("D")+CmtStr.GetSubStr(DocNmPrefixStr.Len(), CmtStr.Len()-1);
BowDocBs->DocNmToDescStrH.DelKey(DocNm);
int NewDId=BowDocBs->DocNmToDescStrH.AddKey(NewDocNm);
IAssert(DId==NewDId);
}
}
SpV->Trunc();
while (Lx.GetSym(syInt, syEoln, syEof)==syEoln){}
}
}
// add missing words
for (int WId=0; WId<=MxWId; WId++){
if (!BowDocBs->IsWId(WId)){
TStr WordStr=TInt::GetStr(WId, "W%d");
int _WId=BowDocBs->AddWordStr(WordStr);
IAssert(WId==_WId);
TInt Fq;
if (WIdToFqH.IsKeyGetDat(WId, Fq)){
BowDocBs->PutWordFq(WId, Fq);
}
}
}
BowDocBs->AssertOk();
return BowDocBs;
}
int FastCorePeripheryGC(PUNGraph& Graph, TIntIntH& out) {
TIntH GroupNodes; // buildup cpntainer of group nodes
int *NNodes = new int[Graph->GetNodes()]; // container of neighbouring nodes
int NNodes_br = 0;
TIntIntH nodes;
TIntIntH nodesIds;
double Z=0;
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) { // Calculate and store the degrees of each node.
int deg = NI.GetDeg();
int id = NI.GetId();
Z += deg;
nodes.AddDat(id,deg);
}
Z = Z/2;
nodes.SortByDat(false); // Then sort the nodes in descending order of degree, to get a list of nodes {v1, v2, . . . , vn}.
int br1=0;
for (THashKeyDatI<TInt,TInt> NI = nodes.BegI(); NI < nodes.EndI(); NI++) {
nodesIds.AddDat(NI.GetKey(),NI.GetKey());
br1++;
}
double Zbest = 99999900000000000;
//int kbest;
//int olddeg;
int br=0;
for (int k=0; k<nodes.Len(); k++) {
if (k<nodes.Len()-1) {
if (nodes[k]==nodes[k+1]) { // go into same deg mode
int kmin=-2;
int knew=-1;
while (kmin < 999999 && kmin !=-1 ) {
int kind=-1;
knew=k;
kmin=999999;
while(nodes[k]==nodes[knew] && knew < nodes.Len()-1) {
int inter = Intersect(Graph->GetNI(nodesIds[knew]),NNodes,NNodes_br);
int deg = nodes[knew];
//if (((((nodes.Len()-NNodes_br)*(nodes.Len()-NNodes_br)))-(nodes.Len()-NNodes_br))/2<(((br*br)-br)/2))
if ((deg-inter)<kmin && !GroupNodes.IsKey(nodesIds[knew]))
{
kmin = deg-inter;
kind = knew;
}
knew++;
}
if (kind!=-1) {
br++;
Z = Z + br - 1 - nodes[kind];
if (Z < (Zbest)) { // or <=
//if (olddeg>nodes[kind])
//olddeg = nodes[kind];
Zbest = Z;
//kbest = br;
int w = nodes[kind];
int id = nodesIds[kind];
GroupNodes.AddDat(id,w);
NNodes[NNodes_br] = id;
NNodes_br++;
}
else {
break;
}
}
}
k=knew-1;
}
else {
br++;
Z = Z + br - 1 - nodes[k];
if (Z < (Zbest)) { // or <=
//if (olddeg>nodes[k])
//olddeg = nodes[k];
Zbest = Z;
//kbest = br;
int w = nodes[k];
int id = nodesIds[k];
GroupNodes.AddDat(id,w);
NNodes[NNodes_br] = id;
NNodes_br++;
}
}
}
else {
br++;
Z = Z + br - 1 - nodes[k];
if (Z < Zbest) { // or <=
//if (olddeg>nodes[k])
//olddeg = nodes[k];
Zbest = Z;
//kbest = br;
int w = nodes[k];
int id = nodesIds[k];
GroupNodes.AddDat(id,w);
NNodes[NNodes_br] = id;
NNodes_br++;
}
}
}
int cp = 0;
br = 0;
for (THashKeyDatI<TInt, TInt> it = nodes.BegI(); !it.IsEnd(); it++) {
if (GroupNodes.IsKey(it.GetKey()))
cp = 1;
else
cp = 0;
out.AddDat(it.GetKey(), cp);
br++;
}
/*for (THashKeyDatI<TInt, TInt> it = GroupNodes.BegI(); it < GroupNodes.EndI(); it++) {
out.AddDat(it.GetKey(), 1);
br++;
}*/
//return kbest;
return GroupNodes.Len();
}