int Intersect(TUNGraph::TNodeI Node, int *NNodes, int NNodes_br){
int br = 0;
int neig;
for (int i = 0; i<Node.GetDeg(); i++)
{
neig = Node.GetNbrNId(i);
for (int j = 0; j<NNodes_br; j++)
{
if (neig == NNodes[j])
{
br++;
j = NNodes_br;
}
}
}
neig = Node.GetId();
for (int j = 0; j<NNodes_br; j++)
{
if (neig == NNodes[j])
{
br++;
j = NNodes_br;
}
}
return br;
}
// Compute the change in likelihood (Delta) if node UID leaves community CID.
double TAGMFit::SeekLeave(const int& UID, const int& CID) {
IAssert(CIDNSetV[CID].IsKey(UID));
IAssert(G->IsNode(UID));
double Delta = 0.0;
TUNGraph::TNodeI NI = G->GetNI(UID);
int NbhsInC = 0;
for (int e = 0; e < NI.GetDeg(); e++) {
const int VID = NI.GetNbrNId(e);
if (! NIDComVH.GetDat(VID).IsKey(CID)) { continue; }
TIntPr SrcDstNIDPr(TMath::Mn(UID,VID), TMath::Mx(UID,VID));
TIntSet& JointCom = EdgeComVH.GetDat(SrcDstNIDPr);
double CurPuv, NewPuv, LambdaSum = SelectLambdaSum(JointCom);
CurPuv = 1 - exp(- LambdaSum);
NewPuv = 1 - exp(- LambdaSum + LambdaV[CID]);
IAssert(JointCom.Len() > 0);
if (JointCom.Len() == 1) {
NewPuv = PNoCom;
}
Delta += (log(NewPuv) - log(CurPuv));
IAssert(!_isnan(Delta));
NbhsInC++;
}
Delta += LambdaV[CID] * (CIDNSetV[CID].Len() - 1 - NbhsInC);
return Delta;
}
double TAGMFast::LikelihoodForOneVar(const TFltV& AlphaKV, const int UID, const int CID, const double& Val) {
TUNGraph::TNodeI UI = G->GetNI(UID);
double L = 0.0, PNoEdge;
int VID = 0;
for (int e = 0; e < UI.GetDeg(); e++) {
VID = UI.GetNbrNId(e);
if (HOVIDSV[UID].IsKey(UI.GetNbrNId(e))) { continue; }
if (! F[VID].IsKey(CID)) {
PNoEdge = AlphaKV[e];
} else {
PNoEdge = AlphaKV[e] * exp (- F[VID].GetDat(CID) * Val);
}
IAssert(PNoEdge <= 1.0 && PNoEdge >= 0.0);
//PNoEdge = PNoEdge >= 1.0 - PNoCom? 1 - PNoCom: PNoEdge;
L += log(1.0 - PNoEdge) + NegWgt * GetCom(VID, CID) * Val;
// += ((PNoEdge * F[VID].GetDat(CID)) / (1.0 - PNoEdge) + NegWgt * F[VID].GetDat(CID));
}
L -= NegWgt * (SumFV[CID] - GetCom(UID, CID)) * Val;
//add regularization
if (RegCoef > 0.0) { //L1
L -= RegCoef * Val;
}
if (RegCoef < 0.0) { //L2
L += RegCoef * Val * Val;
}
return L;
}
THash<TInt, TInt> * choose_seeds (const PUNGraph g, const int num, const int * infection_state, const int infect) {
THash<TInt, TInt> choices;
THash<TInt, TUNGraph::TNode> nodes;
THash<TInt, TInt> * output = new THash<TInt, TInt> ();
TInt weight = 0;
TInt num_total = 0;
for (TUNGraph::TNodeI n = g->BegNI(); n != g->EndNI(); n++) {
//cout << "nodeID: " << n.GetId() << ",\tStatus: " << infection_state[n.GetId () - 1] << endl;
if (infection_state[n.GetId () - 1] != infect) {
weight += n.GetDeg ();
choices.AddDat (num_total, weight);
nodes.AddDat (num_total, n.GetId());
num_total++;
}
}
// TRnd random ((int) time(NULL));
// TRnd random (0);
TInt num_chosen = 0;
while (num_chosen < num) {
TInt choice = my_random.GetUniDevInt (weight);
TUNGraph::TNode node_choice = nodes[find (choice, choices, 0, num_total-1)];
if (!output->IsKey(node_choice.GetId())) {
num_chosen++;
// cout << node_choice.GetId () << "\n";
output->AddDat(node_choice.GetId (), 1);
}
}
return output;
}
double ave_path_length (PUNGraph p) {
TVec<TInt> v;
double tot_lengths = 0.0;
for (TUNGraph::TNodeI n = p->BegNI(); n != p->EndNI(); n++) {
v = v + n.GetId();
}
// cerr << "vlen: " << v.Len() << endl;
TBreathFS<PUNGraph> b(p);
double tot_pairs = 0.0;
while (v.Len () > 0) {
TInt last = v[v.Len()-1];
b.DoBfs (last, true, true);
for (TVec<TInt>::TIter i = v.BegI(); (*i) != last; i++) {
int length;
length = b.GetHops (last, (*i));
if (length == length) {
tot_lengths += length;
tot_pairs += 1;
}
}
// cerr << "tps: " << tot_pairs << ", last: " << last << ", beg: " << v[*(v.BegI())] << endl;
v.Del(v.Len()-1);
}
// cerr << "paths: " << tot_lengths << " " << tot_pairs << " " << (tot_lengths/tot_pairs) << endl;
return tot_lengths / tot_pairs;
}
double TAGMUtil::GetConductance(const PUNGraph& Graph, const TIntSet& CmtyS, const int Edges) {
const int Edges2 = Edges >= 0 ? 2*Edges : Graph->GetEdges();
int Vol = 0, Cut = 0;
double Phi = 0.0;
for (int i = 0; i < CmtyS.Len(); i++) {
if (! Graph->IsNode(CmtyS[i])) {
continue;
}
TUNGraph::TNodeI NI = Graph->GetNI(CmtyS[i]);
for (int e = 0; e < NI.GetOutDeg(); e++) {
if (! CmtyS.IsKey(NI.GetOutNId(e))) {
Cut += 1;
}
}
Vol += NI.GetOutDeg();
}
// get conductance
if (Vol != Edges2) {
if (2 * Vol > Edges2) {
Phi = Cut / double (Edges2 - Vol);
}
else if (Vol == 0) {
Phi = 0.0;
}
else {
Phi = Cut / double(Vol);
}
} else {
if (Vol == Edges2) {
Phi = 1.0;
}
}
return Phi;
}
// Compute the change in likelihood (Delta) if node UID switches from CurCID to NewCID.
double TAGMFit::SeekSwitch(const int& UID, const int& CurCID, const int& NewCID) {
IAssert(! CIDNSetV[NewCID].IsKey(UID));
IAssert(CIDNSetV[CurCID].IsKey(UID));
double Delta = SeekJoin(UID, NewCID) + SeekLeave(UID, CurCID);
//correct only for intersection between new com and current com
TUNGraph::TNodeI NI = G->GetNI(UID);
for (int e = 0; e < NI.GetDeg(); e++) {
const int VID = NI.GetNbrNId(e);
if (! NIDComVH.GetDat(VID).IsKey(CurCID) || ! NIDComVH.GetDat(VID).IsKey(NewCID)) {continue;}
TIntPr SrcDstNIDPr(TMath::Mn(UID,VID), TMath::Mx(UID,VID));
TIntSet& JointCom = EdgeComVH.GetDat(SrcDstNIDPr);
double CurPuv, NewPuvAfterJoin, NewPuvAfterLeave, NewPuvAfterSwitch, LambdaSum = SelectLambdaSum(JointCom);
CurPuv = 1 - exp(- LambdaSum);
NewPuvAfterLeave = 1 - exp(- LambdaSum + LambdaV[CurCID]);
NewPuvAfterJoin = 1 - exp(- LambdaSum - LambdaV[NewCID]);
NewPuvAfterSwitch = 1 - exp(- LambdaSum - LambdaV[NewCID] + LambdaV[CurCID]);
if (JointCom.Len() == 1 || NewPuvAfterLeave == 0.0) {
NewPuvAfterLeave = PNoCom;
}
Delta += (log(NewPuvAfterSwitch) + log(CurPuv) - log(NewPuvAfterLeave) - log(NewPuvAfterJoin));
if (_isnan(Delta)) {
printf("NS:%f C:%f NL:%f NJ:%f PNoCom:%f", NewPuvAfterSwitch, CurPuv, NewPuvAfterLeave, NewPuvAfterJoin, PNoCom.Val);
}
IAssert(!_isnan(Delta));
}
return Delta;
}
void TAGMFit::InitNodeData() {
TSnap::DelSelfEdges(G);
NIDComVH.Gen(G->GetNodes());
for (TUNGraph::TNodeI NI = G->BegNI(); NI < G->EndNI(); NI++) {
NIDComVH.AddDat(NI.GetId());
}
TAGMUtil::GetNodeMembership(NIDComVH, CIDNSetV);
GetEdgeJointCom();
LambdaV.Gen(CIDNSetV.Len());
for (int c = 0; c < CIDNSetV.Len(); c++) {
int MaxE = (CIDNSetV[c].Len()) * (CIDNSetV[c].Len() - 1) / 2;
if (MaxE < 2) {
LambdaV[c] = MaxLambda;
}
else{
LambdaV[c] = -log((double) (MaxE - ComEdgesV[c]) / MaxE);
}
if (LambdaV[c] > MaxLambda) { LambdaV[c] = MaxLambda; }
if (LambdaV[c] < MinLambda) { LambdaV[c] = MinLambda; }
}
NIDCIDPrS.Gen(G->GetNodes() * 10);
for (int c = 0; c < CIDNSetV.Len(); c++) {
for (TIntSet::TIter SI = CIDNSetV[c].BegI(); SI < CIDNSetV[c].EndI(); SI++) {
NIDCIDPrS.AddKey(TIntPr(SI.GetKey(), c));
}
}
}
void TAGMUtil::GetNbhCom(const PUNGraph& Graph, const int NID, TIntSet& NBCmtyS) {
TUNGraph::TNodeI NI = Graph->GetNI(NID);
NBCmtyS.Gen(NI.GetDeg());
NBCmtyS.AddKey(NID);
for (int e = 0; e < NI.GetDeg(); e++) {
NBCmtyS.AddKey(NI.GetNbrNId(e));
}
}
int Intersect(TUNGraph::TNodeI Node, TIntH NNodes){
int br = 0;
for (int i = 0; i<Node.GetDeg(); i++)
{
if (NNodes.IsKey(Node.GetNbrNId(i)))
br++;
}
if (NNodes.IsKey(Node.GetId()))
br++;
return br;
}
double GetGroupDegreeCentr(const PUNGraph& Graph, const PUNGraph& Group) {
int deg;
TIntH NN;
for (TUNGraph::TNodeI NI = Group->BegNI(); NI < Group->EndNI(); NI++) {
deg = Graph->GetNI(NI.GetId()).GetDeg();
for (int i = 0; i<deg; i++) {
if (Group->IsNode(Graph->GetNI(NI.GetId()).GetNbrNId(i)) == 0)
NN.AddDat(Graph->GetNI(NI.GetId()).GetNbrNId(i), NI.GetId());
}
}
return (double)NN.Len();
}
int main(int argc, char* argv[]) {
Env = TEnv(argc, argv, TNotify::StdNotify);
Env.PrepArgs(TStr::Fmt("Node Centrality. build: %s, %s. Time: %s", __TIME__, __DATE__, TExeTm::GetCurTm()));
TExeTm ExeTm;
Try
const TStr InFNm = Env.GetIfArgPrefixStr("-i:", "../as20graph.txt", "Input un/directed graph");
const TStr OutFNm = Env.GetIfArgPrefixStr("-o:", "node_centrality.tab", "Output file");
printf("Loading %s...", InFNm.CStr());
PNGraph Graph = TSnap::LoadEdgeList<PNGraph>(InFNm);
//PNGraph Graph = TSnap::GenRndGnm<PNGraph>(10, 10);
//TGraphViz::Plot(Graph, gvlNeato, InFNm+".gif", InFNm, true);
printf("nodes:%d edges:%d\n", Graph->GetNodes(), Graph->GetEdges());
PUNGraph UGraph = TSnap::ConvertGraph<PUNGraph>(Graph); // undirected version of the graph
TIntFltH BtwH, EigH, PRankH, CcfH, CloseH, HubH, AuthH;
//printf("Computing...\n");
printf("Treat graph as DIRECTED: ");
printf(" PageRank... "); TSnap::GetPageRank(Graph, PRankH, 0.85);
printf(" Hubs&Authorities..."); TSnap::GetHits(Graph, HubH, AuthH);
printf("\nTreat graph as UNDIRECTED: ");
printf(" Eigenvector..."); TSnap::GetEigenVectorCentr(UGraph, EigH);
printf(" Clustering..."); TSnap::GetNodeClustCf(UGraph, CcfH);
printf(" Betweenness (SLOW!)..."); TSnap::GetBetweennessCentr(UGraph, BtwH, 1.0);
printf(" Constraint (SLOW!)..."); TNetConstraint<PUNGraph> NetC(UGraph, true);
printf(" Closeness (SLOW!)...");
for (TUNGraph::TNodeI NI = UGraph->BegNI(); NI < UGraph->EndNI(); NI++) {
const int NId = NI.GetId();
CloseH.AddDat(NId, TSnap::GetClosenessCentr<PUNGraph>(UGraph, NId, false));
}
printf("\nDONE! saving...");
FILE *F = fopen(OutFNm.CStr(), "wt");
fprintf(F,"#Network: %s\n", InFNm.CStr());
fprintf(F,"#Nodes: %d\tEdges: %d\n", Graph->GetNodes(), Graph->GetEdges());
fprintf(F,"#NodeId\tDegree\tCloseness\tBetweennes\tEigenVector\tNetworkConstraint\tClusteringCoefficient\tPageRank\tHubScore\tAuthorityScore\n");
for (TUNGraph::TNodeI NI = UGraph->BegNI(); NI < UGraph->EndNI(); NI++) {
const int NId = NI.GetId();
const double DegCentr = UGraph->GetNI(NId).GetDeg();
const double CloCentr = CloseH.GetDat(NId);
const double BtwCentr = BtwH.GetDat(NId);
const double EigCentr = EigH.GetDat(NId);
const double Constraint = NetC.GetNodeC(NId);
const double ClustCf = CcfH.GetDat(NId);
const double PgrCentr = PRankH.GetDat(NId);
const double HubCentr = HubH.GetDat(NId);
const double AuthCentr = AuthH.GetDat(NId);
fprintf(F, "%d\t%f\t%f\t%f\t%f\t%f\t%f\t%f\t%f\t%f\n", NId,
DegCentr, CloCentr, BtwCentr, EigCentr, Constraint, ClustCf, PgrCentr, HubCentr, AuthCentr);
}
fclose(F);
Catch
printf("\nrun time: %s (%s)\n", ExeTm.GetTmStr(), TSecTm::GetCurTm().GetTmStr().CStr());
return 0;
}
// After MCMC, NID joins community CID.
void TAGMFit::JoinCom(const int& NID, const int& JoinCID) {
TUNGraph::TNodeI NI = G->GetNI(NID);
for (int e = 0; e < NI.GetDeg(); e++) {
int VID = NI.GetNbrNId(e);
if (NIDComVH.GetDat(VID).IsKey(JoinCID)) {
TIntPr SrcDstNIDPr = TIntPr(TMath::Mn(NID,VID), TMath::Mx(NID,VID));
EdgeComVH.GetDat(SrcDstNIDPr).AddKey(JoinCID);
ComEdgesV[JoinCID]++;
}
}
CIDNSetV[JoinCID].AddKey(NID);
NIDComVH.GetDat(NID).AddKey(JoinCID);
NIDCIDPrS.AddKey(TIntPr(NID, JoinCID));
}
void TAGMFit::RandomInit(const int& MaxK) {
CIDNSetV.Clr();
for (int c = 0; c < MaxK; c++) {
CIDNSetV.Add();
int NC = Rnd.GetUniDevInt(G -> GetNodes());
TUNGraph::TNodeI NI = G -> GetRndNI();
CIDNSetV.Last().AddKey(NI.GetId());
for (int v = 0; v < NC; v++) {
NI = G->GetNI(NI.GetNbrNId(Rnd.GetUniDevInt(NI.GetDeg())));
CIDNSetV.Last().AddKey(NI.GetId());
}
}
InitNodeData();
SetDefaultPNoCom();
}
static double CmtyCMN(const PUNGraph& Graph, TCnComV& CmtyV) {
TCNMQMatrix QMatrix(Graph);
// maximize modularity
while (QMatrix.MergeBestQ()) { }
// reconstruct communities
THash<TInt, TIntV> IdCmtyH;
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
IdCmtyH.AddDat(QMatrix.CmtyIdUF.Find(NI.GetId())).Add(NI.GetId());
}
CmtyV.Gen(IdCmtyH.Len());
for (int j = 0; j < IdCmtyH.Len(); j++) {
CmtyV[j].NIdV.Swap(IdCmtyH[j]);
}
return QMatrix.Q;
}
// Connected components of a graph define clusters
// OutDegH and OrigEdges stores node degrees and number of edges in the original graph
double _GirvanNewmanGetModularity(const PUNGraph& G, const TIntH& OutDegH, const int& OrigEdges, TCnComV& CnComV) {
TSnap::GetWccs(G, CnComV); // get communities
double Mod = 0;
for (int c = 0; c < CnComV.Len(); c++) {
const TIntV& NIdV = CnComV[c]();
double EIn=0, EEIn=0;
for (int i = 0; i < NIdV.Len(); i++) {
TUNGraph::TNodeI NI = G->GetNI(NIdV[i]);
EIn += NI.GetOutDeg();
EEIn += OutDegH.GetDat(NIdV[i]);
}
Mod += (EIn-EEIn*EEIn/(2.0*OrigEdges));
}
if (Mod == 0) { return 0; }
else { return Mod/(2.0*OrigEdges); }
}
void GetEigenVectorCentr(const PUNGraph& Graph, TIntFltH& NIdEigenH, const double& Eps, const int& MaxIter) {
const int NNodes = Graph->GetNodes();
NIdEigenH.Gen(NNodes);
// initialize vector values
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
NIdEigenH.AddDat(NI.GetId(), 1.0 / NNodes);
IAssert(NI.GetId() == NIdEigenH.GetKey(NIdEigenH.Len() - 1));
}
TFltV TmpV(NNodes);
for (int iter = 0; iter < MaxIter; iter++) {
int j = 0;
// add neighbor values
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++, j++) {
TmpV[j] = 0;
for (int e = 0; e < NI.GetOutDeg(); e++) {
TmpV[j] += NIdEigenH.GetDat(NI.GetOutNId(e));
}
}
// normalize
double sum = 0;
for (int i = 0; i < TmpV.Len(); i++) {
sum += (TmpV[i] * TmpV[i]);
}
sum = sqrt(sum);
for (int i = 0; i < TmpV.Len(); i++) {
TmpV[i] /= sum;
}
// compute difference
double diff = 0.0;
j = 0;
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++, j++) {
diff += fabs(NIdEigenH.GetDat(NI.GetId()) - TmpV[j]);
}
// set new values
j = 0;
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++, j++) {
NIdEigenH.AddDat(NI.GetId(), TmpV[j]);
}
if (diff < Eps) {
break;
}
}
}
// network cascade: add spurious edges
// for more details see "Correcting for Missing Data in Information Cascades" by E. Sadikov, M. Medina, J. Leskovec, H. Garcia-Molina. WSDM, 2011
PNGraph AddSpuriousEdges(const PUNGraph& Graph, const PNGraph& Casc, TIntH NIdTmH) {
TIntPrV EdgeV;
for (TNGraph::TNodeI NI = Casc->BegNI(); NI < Casc->EndNI(); NI++) {
TUNGraph::TNodeI GNI = Graph->GetNI(NI.GetId());
const int Tm = NIdTmH.GetDat(NI.GetId());
for (int i=0,j=0; i < GNI.GetOutDeg(); i++) {
const int Dst = GNI.GetOutNId(i);
if (NIdTmH.IsKey(Dst) && Tm<NIdTmH.GetDat(Dst) && ! NI.IsNbhNId(Dst)) {
EdgeV.Add(TIntPr(GNI.GetId(), Dst)); }
}
}
PNGraph NetCasc = TNGraph::New();
*NetCasc = *Casc;
for (int e = 0; e < EdgeV.Len(); e++) {
NetCasc->AddEdge(EdgeV[e].Val1, EdgeV[e].Val2); }
return NetCasc;
}
double GetDegreeCentralization(const PUNGraph& Graph) {
int MaxDeg = -1;
int N = Graph->GetNodes();
int Sum = 0;
if (Graph->GetNodes() > 1 && (double(N - 2.0)*double(N - 1)) > 0) {
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
int deg = NI.GetDeg();
if (deg > MaxDeg) {
MaxDeg = deg;
}
}
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
Sum += MaxDeg - NI.GetDeg();
}
return double(Sum) / (double(N - 2.0)*double(N - 1));
}
else { return 0.0; }
}
int TCliqueOverlap::MaxNbrsInCANDNodeId(const THashSet<TInt>& SUBG, const THashSet<TInt>& CAND) const{
int id = -1;
int maxIntersection = -1;
//
for (THashSetKeyI<TInt> it=SUBG.BegI(); it<SUBG.EndI(); it++) {
int nId = it.GetKey();
TUNGraph::TNodeI nIt = m_G->GetNI(nId);
int deg = nIt.GetDeg();
//
int curIntersection = 0;
for (int i=0; i<deg; i++) {
int nbrId = nIt.GetNbrNId(i);
if (CAND.IsKey(nbrId)) curIntersection++;
}
//
if (maxIntersection < curIntersection) { maxIntersection=curIntersection; id=nId; }
}
return id;
}
double GetGroupDegreeCentr(const PUNGraph& Graph, const TIntH& GroupNodes) {
int deg;
TIntH NN;
TIntH GroupNodes1;
for (THashKeyDatI<TInt, TInt> NI = GroupNodes.BegI(); NI < GroupNodes.EndI(); NI++)
GroupNodes1.AddDat(NI.GetDat(), NI.GetDat());
for (THashKeyDatI<TInt, TInt> NI = GroupNodes1.BegI(); NI < GroupNodes1.EndI(); NI++){
TUNGraph::TNodeI node = Graph->GetNI(NI.GetKey());
deg = node.GetDeg();
for (int j = 0; j < deg; j++){
if (GroupNodes1.IsKey(node.GetNbrNId(j)) == 0 && NN.IsKey(node.GetNbrNId(j)) == 0)
NN.AddDat(node.GetNbrNId(j), NI.GetKey());
}
}
return (double)NN.Len();
}
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;
}
int Intersect1(TUNGraph::TNodeI Node, TStr NNodes){
int br = 0;
for (int i = 0; i<Node.GetDeg(); i++)
{
TInt digi = Node.GetNbrNId(i);
TStr buf = "";
buf = digi.GetStr();
if (NNodes.SearchStr(buf.CStr()) != -1)
br++;
}
TInt digi = Node.GetId();
TStr buf = digi.GetStr();
if (NNodes.SearchStr(buf.CStr()) != -1)
br++;
return br;
}
// For each (u, v) in edges, precompute C_uv (the set of communities u and v share).
void TAGMFit::GetEdgeJointCom() {
ComEdgesV.Gen(CIDNSetV.Len());
EdgeComVH.Gen(G->GetEdges());
for (TUNGraph::TNodeI SrcNI = G->BegNI(); SrcNI < G->EndNI(); SrcNI++) {
int SrcNID = SrcNI.GetId();
for (int v = 0; v < SrcNI.GetDeg(); v++) {
int DstNID = SrcNI.GetNbrNId(v);
if (SrcNID >= DstNID) { continue; }
TIntSet JointCom;
IAssert(NIDComVH.IsKey(SrcNID));
IAssert(NIDComVH.IsKey(DstNID));
TAGMUtil::GetIntersection(NIDComVH.GetDat(SrcNID), NIDComVH.GetDat(DstNID), JointCom);
EdgeComVH.AddDat(TIntPr(SrcNID,DstNID),JointCom);
for (int k = 0; k < JointCom.Len(); k++) {
ComEdgesV[JointCom[k]]++;
}
}
}
IAssert(EdgeComVH.Len() == G->GetEdges());
}
double TAGMFast::HessianForOneVar(const TFltV& AlphaKV, const int UID, const int CID, const double& Val) {
TUNGraph::TNodeI UI = G->GetNI(UID);
double H = 0.0, PNoEdge;
int VID = 0;
for (int e = 0; e < UI.GetDeg(); e++) {
VID = UI.GetNbrNId(e);
if (HOVIDSV[UID].IsKey(UI.GetNbrNId(e))) { continue; }
if (! F[VID].IsKey(CID)) { continue; }
PNoEdge = AlphaKV[e] * exp (- F[VID].GetDat(CID) * Val);
IAssert(PNoEdge <= 1.0 && PNoEdge >= 0.0);
//PNoEdge = PNoEdge == 1.0? 1 - PNoCom: PNoEdge;
H += (- PNoEdge * F[VID].GetDat(CID) * F[VID].GetDat(CID)) / (1.0 - PNoEdge) / (1.0 - PNoEdge);
}
//add regularization
if (RegCoef < 0.0) { //L2
H += 2 * RegCoef;
}
IAssert (H <= 0.0);
return H;
}
// renumber node ids to 0...N-1
PUNGraph GetSubGraph(const PUNGraph& Graph, const TIntV& NIdV, const bool& RenumberNodes) {
if (! RenumberNodes) { return TSnap::GetSubGraph(Graph, NIdV); }
PUNGraph NewGraphPt = TUNGraph::New();
TUNGraph& NewGraph = *NewGraphPt;
NewGraph.Reserve(NIdV.Len(), -1);
TIntSet NIdSet(NIdV.Len());
for (int n = 0; n < NIdV.Len(); n++) {
NewGraph.AddNode(n);
NIdSet.AddKey(NIdV[n]);
}
for (int n = 0; n < NIdV.Len(); n++) {
const TUNGraph::TNodeI NI = Graph->GetNI(NIdV[n]);
const int SrcNId = NIdSet.GetKeyId(NI.GetId());
for (int edge = 0; edge < NI.GetDeg(); edge++) {
const int OutNId = NIdSet.GetKeyId(NI.GetNbhNId(edge));
if (NewGraph.IsNode(OutNId)) {
NewGraph.AddEdge(SrcNId, OutNId); }
}
}
return NewGraphPt;
}
void GetEigenVectorCentr(const PUNGraph& Graph, TIntFltH& EigenH, const double& Eps, const int& MaxIter) {
const int NNodes = Graph->GetNodes();
EigenH.Gen(NNodes);
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
EigenH.AddDat(NI.GetId(), 1.0/NNodes);
IAssert(NI.GetId() == EigenH.GetKey(EigenH.Len()-1));
}
TFltV TmpV(NNodes);
double diff = TFlt::Mx;
for (int iter = 0; iter < MaxIter; iter++) {
int j = 0;
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++, j++) {
TmpV[j] = 0;
for (int e = 0; e < NI.GetOutDeg(); e++) {
TmpV[j] += EigenH.GetDat(NI.GetOutNId(e)); }
}
double sum = 0;
for (int i = 0; i < TmpV.Len(); i++) {
EigenH[i] = TmpV[i];
sum += EigenH[i];
}
for (int i = 0; i < EigenH.Len(); i++) {
EigenH[i] /= sum; }
if (fabs(diff-sum) < Eps) { break; }
//printf("\tdiff:%f\tsum:%f\n", fabs(diff-sum), sum);
diff = sum;
}
}
// Maximum modularity clustering by Girvan-Newman algorithm (slow)
// Girvan M. and Newman M. E. J., Community structure in social and biological networks, Proc. Natl. Acad. Sci. USA 99, 7821–7826 (2002)
double CommunityGirvanNewman(PUNGraph& Graph, TCnComV& CmtyV) {
TIntH OutDegH;
const int NEdges = Graph->GetEdges();
for (TUNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
OutDegH.AddDat(NI.GetId(), NI.GetOutDeg());
}
double BestQ = -1; // modularity
TCnComV CurCmtyV;
CmtyV.Clr();
TIntV Cmty1, Cmty2;
while (true) {
CmtyGirvanNewmanStep(Graph, Cmty1, Cmty2);
const double Q = _GirvanNewmanGetModularity(Graph, OutDegH, NEdges, CurCmtyV);
//printf("current modularity: %f\n", Q);
if (Q > BestQ) {
BestQ = Q;
CmtyV.Swap(CurCmtyV);
}
if (Cmty1.Len()==0 || Cmty2.Len() == 0) { break; }
}
return BestQ;
}
int Intersect(TUNGraph::TNodeI Node, TStr NNodes){
int br = 0;
TInt digi = -1;
TStr buf = "";
for (int i = 0; i<Node.GetDeg(); i++)
{
digi = Node.GetNbrNId(i);
TStr buf = digi.GetStr();
if (NNodes.IsStrIn(buf.CStr()))
br++;
}
digi = Node.GetId();
buf = digi.GetStr();
if (NNodes.IsStrIn(buf.CStr()))
br++;
return br;
}
double TAGMFast::GradientForOneVar(const TFltV& AlphaKV, const int UID, const int CID, const double& Val) {
TUNGraph::TNodeI UI = G->GetNI(UID);
double Grad = 0.0, PNoEdge;
int VID = 0;
for (int e = 0; e < UI.GetDeg(); e++) {
VID = UI.GetNbrNId(e);
if (HOVIDSV[UID].IsKey(UI.GetNbrNId(e))) { continue; }
if (! F[VID].IsKey(CID)) { continue; }
PNoEdge = AlphaKV[e] * exp (- F[VID].GetDat(CID) * Val);
IAssert(PNoEdge <= 1.0 && PNoEdge >= 0.0);
//PNoEdge = PNoEdge >= 1.0 - PNoCom? 1 - PNoCom: PNoEdge;
Grad += ((PNoEdge * F[VID].GetDat(CID)) / (1.0 - PNoEdge) + NegWgt * F[VID].GetDat(CID));
}
Grad -= NegWgt * (SumFV[CID] - GetCom(UID, CID));
//add regularization
if (RegCoef > 0.0) { //L1
Grad -= RegCoef;
}
if (RegCoef < 0.0) { //L2
Grad += 2 * RegCoef * Val;
}
return Grad;
}