/////////////////////////////////////////////////
// Trawling the web for emerging communities
// graph, left points to right
TTrawling::TTrawling(const PNGraph& Graph, const int& MinSupport) : MinSup(MinSupport) {
TIntH ItemCntH;
for (TNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
IAssert(NI.GetOutDeg()==0 || NI.GetInDeg()==0); // edges only point from left to right
if (NI.GetOutDeg()==0) { continue; }
for (int e = 0; e < NI.GetOutDeg(); e++) {
ItemCntH.AddDat(NI.GetOutNId(e)) += 1;
}
}
TIntV RightV;
for (TNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
IAssert(NI.GetOutDeg()==0 || NI.GetInDeg()==0); // edges only point from left to right
if (NI.GetOutDeg()==0) { continue; }
RightV.Clr(false);
for (int e = 0; e < NI.GetOutDeg(); e++) {
const int itm = NI.GetOutNId(e);
// only include items that already are above minimum support
if (ItemCntH.GetDat(itm) >= MinSup) {
RightV.Add(itm); }
}
if (! RightV.Empty()) {
NIdSetH.AddDat(NI.GetId(), RightV);
}
}
//
for (int n = 0; n < NIdSetH.Len(); n++) {
const TIntV& Set = NIdSetH[n];
for (int s = 0; s < Set.Len(); s++) {
SetNIdH.AddDat(Set[s]).Add(n);
}
}
}
double DirectedModularity(PNGraph& graph, std::vector<int>& communities) {
if (graph->GetNodes() != communities.size()) {
throw std::logic_error("Number of nodes does not match community size.");
}
int num_edges = graph->GetEdges();
double score = 0.0;
int num_unique = 10;
std::map<int, double> outdeg_sums;
std::map<int, double> indeg_sums;
for (TNGraph::TNodeI node = graph->BegNI(); node < graph->EndNI(); node++) {
int comm = communities[node.GetId()];
outdeg_sums[comm] += node.GetOutDeg();
indeg_sums[comm] += node.GetInDeg();
}
for (auto& kv : outdeg_sums) {
score -= (kv.second / num_edges) * indeg_sums[kv.first];
}
for (TNGraph::TNodeI node = graph->BegNI(); node < graph->EndNI(); node++) {
int node_ID = node.GetId();
for (int e = 0; e < node.GetOutDeg(); ++e) {
int nbr = node.GetOutNId(e);
if (communities[node_ID] == communities[nbr]) {
score += 1.0;
}
}
}
return score / num_edges;
}
void TSubGraphsEnum::RecurBfs1(const int& NId, const int& Depth) {
if (Depth == 0) {
TIntPrV EdgeV;
EdgeH.GetKeyV(EdgeV);
EdgeV.Sort();
SgV.Add(EdgeV);
return;
}
const TNGraph::TNodeI NI = NGraph ->GetNI(NId);
for (int e = 0; e < NI.GetOutDeg(); e++) {
const TIntPr Edge(NId, NI.GetOutNId(e));
if (! EdgeH.IsKey(Edge)) {
EdgeH.AddKey(Edge);
RecurBfs1(NI.GetOutNId(e), Depth-1);
EdgeH.DelKey(Edge);
}
}
for (int e = 0; e < NI.GetInDeg(); e++) {
const TIntPr Edge(NI.GetInNId(e), NId);
if (! EdgeH.IsKey(Edge)) {
EdgeH.AddKey(Edge);
RecurBfs1(NI.GetInNId(e), Depth-1);
EdgeH.DelKey(Edge);
}
}
}
void TGraphCascade::PruneGraph() {
// iterate over nodes
int Nodes = NodeNmIdH.Len();
TIntV NodeIdV; NodeNmIdH.GetDatV(NodeIdV);
TStrV NodeNmV; NodeNmIdH.GetKeyV(NodeNmV);
for (int NodeN = 0; NodeN < Nodes; NodeN++) {
int NodeId = NodeIdV[NodeN];
if (!EnabledNodeIdH.IsKey(NodeId)) {
// if a node is not enabled:
// - connect its parents to its children
TNGraph::TNodeI NI = Graph.GetNI(NodeId);
for (int ParentN = 0; ParentN < NI.GetInDeg(); ParentN++) {
for (int ChildN = 0; ChildN < NI.GetOutDeg(); ChildN++) {
if (!Graph.IsEdge(NI.GetInNId(ParentN), NI.GetOutNId(ChildN))) {
Graph.AddEdge(NI.GetInNId(ParentN), NI.GetOutNId(ChildN));
}
}
}
//printf("deleting node %s %d\n", NodeNmV[NodeN].CStr(), NodeId);
// - delete it (deletes edges)
Graph.DelNode(NodeId);
}
}
// generate search sequence from sinks to sources
TopologicalSort(NIdSweep);
//Print(NIdSweep);
}
// burn each link independently (forward with FwdBurnProb, backward with BckBurnProb)
void TForestFire::BurnExpFire() {
const double OldFwdBurnProb = FwdBurnProb;
const double OldBckBurnProb = BckBurnProb;
const int NInfect = InfectNIdV.Len();
const TNGraph& G = *Graph;
TIntH BurnedNIdH; // burned nodes
TIntV BurningNIdV = InfectNIdV; // currently burning nodes
TIntV NewBurnedNIdV; // nodes newly burned in current step
bool HasAliveNbrs; // has unburned neighbors
int NBurned = NInfect, NDiedFire=0;
for (int i = 0; i < InfectNIdV.Len(); i++) {
BurnedNIdH.AddDat(InfectNIdV[i]); }
NBurnedTmV.Clr(false); NBurningTmV.Clr(false); NewBurnedTmV.Clr(false);
for (int time = 0; ; time++) {
NewBurnedNIdV.Clr(false);
// for each burning node
for (int node = 0; node < BurningNIdV.Len(); node++) {
const int& BurningNId = BurningNIdV[node];
const TNGraph::TNodeI Node = G.GetNI(BurningNId);
HasAliveNbrs = false;
NDiedFire = 0;
// burn forward links (out-links)
for (int e = 0; e < Node.GetOutDeg(); e++) {
const int OutNId = Node.GetOutNId(e);
if (! BurnedNIdH.IsKey(OutNId)) { // not yet burned
HasAliveNbrs = true;
if (Rnd.GetUniDev() < FwdBurnProb) {
BurnedNIdH.AddDat(OutNId); NewBurnedNIdV.Add(OutNId); NBurned++; }
}
}
// burn backward links (in-links)
if (BckBurnProb > 0.0) {
for (int e = 0; e < Node.GetInDeg(); e++) {
const int InNId = Node.GetInNId(e);
if (! BurnedNIdH.IsKey(InNId)) { // not yet burned
HasAliveNbrs = true;
if (Rnd.GetUniDev() < BckBurnProb) {
BurnedNIdH.AddDat(InNId); NewBurnedNIdV.Add(InNId); NBurned++; }
}
}
}
if (! HasAliveNbrs) { NDiedFire++; }
}
NBurnedTmV.Add(NBurned);
NBurningTmV.Add(BurningNIdV.Len() - NDiedFire);
NewBurnedTmV.Add(NewBurnedNIdV.Len());
//BurningNIdV.AddV(NewBurnedNIdV); // node is burning eternally
BurningNIdV.Swap(NewBurnedNIdV); // node is burning just 1 time step
if (BurningNIdV.Empty()) break;
FwdBurnProb = FwdBurnProb * ProbDecay;
BckBurnProb = BckBurnProb * ProbDecay;
}
BurnedNIdV.Gen(BurnedNIdH.Len(), 0);
for (int i = 0; i < BurnedNIdH.Len(); i++) {
BurnedNIdV.Add(BurnedNIdH.GetKey(i)); }
FwdBurnProb = OldFwdBurnProb;
BckBurnProb = OldBckBurnProb;
}
void TempMotifCounter::GetAllNeighbors(int node, TIntV& nbrs) {
nbrs = TIntV();
TNGraph::TNodeI NI = static_graph_->GetNI(node);
for (int i = 0; i < NI.GetOutDeg(); i++) { nbrs.Add(NI.GetOutNId(i)); }
for (int i = 0; i < NI.GetInDeg(); i++) {
int nbr = NI.GetInNId(i);
if (!NI.IsOutNId(nbr)) { nbrs.Add(nbr); }
}
}
uint64 TGraphCascade::SampleNodeTimestamp(const int& NodeId, const uint64& Time, const int& SampleN) {
if (Timestamps.GetDat(NodeId) > 0) {
return Timestamps.GetDat(NodeId);
} else if (Sample.GetDat(NodeId).Len() > SampleN) {
return Sample.GetDat(NodeId)[SampleN];
}
TNGraph::TNodeI NI = Graph.GetNI(NodeId);
int Parents = NI.GetInDeg();
if (Parents == 0) {
throw TExcept::New("Root node has not been observed yet - cannot run simulation");
}
uint64 MaxParentTime = 0;
for (int ParentN = 0; ParentN < Parents; ParentN++) {
int ParentId = NI.GetInNId(ParentN);
// check if parent was observed or has samples available
if (Timestamps.GetDat(ParentId) == 0 && Sample.GetDat(ParentId).Empty()) {
throw TExcept::New("Parent node unobserved and sample not available! Missing data or bad graph");
}
// get SampleN from each parent
uint64 ParentTime = SampleNodeTimestamp(ParentId, Time, SampleN);
// update max
MaxParentTime = (MaxParentTime > ParentTime) ? MaxParentTime : ParentTime;
}
int TimeDiff = (int)(((int64)(Time)-(int64)(MaxParentTime)) / TimeUnit);
int MinDuration = MAX(TimeDiff, 0);
if (!CDF.IsKey(NodeId)) {
// model is missing and the node has not been observed
// this probably indicates that the node is always missing
// model its duration as 0, so it should have been observed after
// its last parent
return MaxParentTime;
}
TFltV& NodeCDF = CDF.GetDat(NodeId);
int MaxDuration = NodeCDF.Len();
if (TimeDiff > MaxDuration) { return Time + 1; } // out of model bounds
double CDFRemain = MinDuration == 0 ? 1.0 : (1.0 - NodeCDF[MinDuration - 1]);
if (CDFRemain < 1e-16) { return Time + 1; } // numerically out of model bounds
// inverse cdf sample, conditioned on elapsed time
double Samp = (1.0 - CDFRemain) + Rnd.GetUniDev() * CDFRemain;
// find the first CDF bin that exceeds Samp
// SPEEDUP possible with bisection
int BinIdx = MaxDuration;
for (int BinN = MinDuration; BinN < MaxDuration; BinN++) {
if (NodeCDF[BinN] >= Samp) {
BinIdx = BinN;
break;
}
}
// compute time
return MaxParentTime + (uint64)((BinIdx)* TimeUnit);
}
int main(int argc, char* argv[]) {
// create a graph and save it
{ PNGraph Graph = TNGraph::New();
for (int i = 0; i < 10; i++) {
Graph->AddNode(i); }
for (int i = 0; i < 10; i++) {
Graph->AddEdge(i, TInt::Rnd.GetUniDevInt(10)); }
TSnap::SaveEdgeList(Graph, "graph.txt", "Edge list format"); }
// load a graph
PNGraph Graph;
Graph = TSnap::LoadEdgeList<PNGraph>("graph.txt", 0, 1);
// traverse nodes
for (TNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
printf("NodeId: %d, InDegree: %d, OutDegree: %d\n", NI.GetId(), NI.GetInDeg(), NI.GetOutDeg());
printf("OutNodes: ");
for (int e = 0; e < NI.GetOutDeg(); e++) { printf(" %d", NI.GetOutNId(e)); }
printf("\nInNodes: ");
for (int e = 0; e < NI.GetInDeg(); e++) { printf(" %d", NI.GetInNId(e)); }
printf("\n\n");
}
// graph statistic
TSnap::PrintInfo(Graph, "Graph info");
PNGraph MxWcc = TSnap::GetMxWcc(Graph);
TSnap::PrintInfo(MxWcc, "Largest Weakly connected component");
// random graph
PNGraph RndGraph = TSnap::GenRndGnm<PNGraph>(100, 1000);
TGStat GraphStat(RndGraph, TSecTm(1), TGStat::AllStat(), "Gnm graph");
GraphStat.PlotAll("RndGraph", "Random graph on 1000 nodes");
// Forest Fire graph
{ TFfGGen ForestFire(false, 1, 0.35, 0.30, 1.0, 0.0, 0.0);
ForestFire.GenGraph(100);
PNGraph FfGraph = ForestFire.GetGraph(); }
// network
TPt<TNodeEDatNet<TStr, TStr> > Net = TNodeEDatNet<TStr, TStr>::New();
Net->AddNode(0, "zero");
Net->AddNode(1, "one");
Net->AddEdge(0, 1, "zero to one");
return 0;
}
// improved version
void GetMergeSortedV1(TIntV& NeighbourV, TNGraph::TNodeI NI) {
int j = 0;
int k = 0;
int prev = -1;
int indeg = NI.GetInDeg();
int outdeg = NI.GetOutDeg();
//while (j < NI.GetInDeg() && k < NI.GetOutDeg()) {
if (indeg > 0 && outdeg > 0) {
int v1 = NI.GetInNId(j);
int v2 = NI.GetOutNId(k);
while (1) {
if (v1 <= v2) {
if (prev != v1) {
NeighbourV.Add(v1);
prev = v1;
}
j += 1;
if (j >= indeg) {
break;
}
v1 = NI.GetInNId(j);
} else {
if (prev != v2) {
NeighbourV.Add(v2);
prev = v2;
}
k += 1;
if (k >= outdeg) {
break;
}
v2 = NI.GetOutNId(k);
}
}
}
while (j < indeg) {
int v = NI.GetInNId(j);
if (prev != v) {
NeighbourV.Add(v);
prev = v;
}
j += 1;
}
while (k < outdeg) {
int v = NI.GetOutNId(k);
if (prev != v) {
NeighbourV.Add(v);
prev = v;
}
k += 1;
}
}
void MergeNbrs(TIntV* NeighbourV, TIntV* list1, TNGraph::TNodeI NI2) {
int j = 0;
int k = 0;
int prev = -1;
int lenA = list1->Len();
int lenB = NI2.GetInDeg();
if (lenA > 0 && lenB > 0) {
int v1 = (*list1)[j];
int v2 = NI2.GetInNId(k);
while (1) {
if (v1 <= v2) {
if (prev != v1) {
NeighbourV->Add(v1);
prev = v1;
}
j += 1;
if (j >= lenA) {
break;
}
v1 = (*list1)[j];
} else {
if (prev != v2) {
NeighbourV->Add(v2);
prev = v2;
}
k += 1;
if (k >= lenB) {
break;
}
v2 = NI2.GetInNId(k);
}
}
}
while (j < lenA) {
int v = (*list1)[j];
if (prev != v) {
NeighbourV->Add(v);
prev = v;
}
j += 1;
}
while (k < lenB) {
int v = NI2.GetInNId(k);
if (prev != v) {
NeighbourV->Add(v);
prev = v;
}
k += 1;
}
}
void TGraphKey::TakeSig(const PNGraph& Graph, const int& MnSvdGraph, const int& MxSvdGraph) {
const int Edges = Graph->GetEdges();
Nodes = Graph->GetNodes();
VariantId = 0;
SigV.Gen(2+Nodes, 0);
// degree sequence
TIntPrV DegV(Nodes, 0);
for (TNGraph::TNodeI NodeI = Graph->BegNI(); NodeI < Graph->EndNI(); NodeI++) {
DegV.Add(TIntPr(NodeI.GetInDeg(), NodeI.GetOutDeg()));
}
DegV.Sort(false);
SigV.Add(TFlt(Nodes));
SigV.Add(TFlt(Edges));
for (int i = 0; i < DegV.Len(); i++) {
SigV.Add(DegV[i].Val1());
SigV.Add(DegV[i].Val2());
}
// singular values signature
// it turns out that it is cheaper to do brute force isomorphism
// checking than to calculate SVD and then check isomorphism
if (Nodes >= MnSvdGraph && Nodes < MxSvdGraph) {
// perform full SVD
TFltVV AdjMtx(Nodes+1, Nodes+1);
TFltV SngValV;
TFltVV LSingV, RSingV;
TIntH NodeIdH;
// create adjecency matrix
for (TNGraph::TNodeI NodeI = Graph->BegNI(); NodeI < Graph->EndNI(); NodeI++) {
NodeIdH.AddKey(NodeI.GetId());
}
for (TNGraph::TNodeI NodeI = Graph->BegNI(); NodeI < Graph->EndNI(); NodeI++) {
const int NodeId = NodeIdH.GetKeyId(NodeI.GetId()) + 1;
for (int e = 0; e < NodeI.GetOutDeg(); e++) {
const int DstNId = NodeIdH.GetKeyId(NodeI.GetOutNId(e)) + 1; // no self edges
if (NodeId != DstNId) AdjMtx.At(NodeId, DstNId) = 1;
}
}
try { // can fail to converge but results seem to be good
TSvd::Svd(AdjMtx, LSingV, SngValV, RSingV);
} catch(...) {
printf("\n***No SVD convergence: G(%d, %d): SngValV.Len():%d\n", Nodes(), Graph->GetEdges(), SngValV.Len());
}
// round singular values
SngValV.Sort(false);
for (int i = 0; i < SngValV.Len(); i++) {
SigV.Add(TMath::Round(SngValV[i], RoundTo));
}
}
//printf("SIG:\n"); for (int i = 0; i < SigV.Len(); i++) { printf("\t%f\n", SigV[i]); }
SigV.Pack();
}
// initial version
void GetMergeSortedV(TIntV& NeighbourV, TNGraph::TNodeI NI) {
int ind, j, k;
ind = j = k = 0;
while (j < NI.GetInDeg() && k < NI.GetOutDeg()) {
int v1 = NI.GetInNId(j);
int v2 = NI.GetOutNId(k);
if (v1 <= v2) {
if ((ind == 0) || (NeighbourV[ind-1] != v1)) {
NeighbourV.Add(v1);
ind += 1;
}
j += 1;
}
else {
if ((ind == 0) || (NeighbourV[ind-1] != v2)) {
NeighbourV.Add(v2);
ind += 1;
}
k += 1;
}
}
while (j < NI.GetInDeg()) {
int v = NI.GetInNId(j);
if ((ind == 0) || (NeighbourV[ind-1] != v)) {
NeighbourV.Add(v);
ind += 1;
}
j += 1;
}
while (k < NI.GetOutDeg()) {
int v = NI.GetOutNId(k);
if ((ind == 0) || (NeighbourV[ind-1] != v)) {
NeighbourV.Add(v);
ind += 1;
}
k += 1;
}
}
// Rok #5
void GetMergeSortedV(TIntV& NeighbourV, TNGraph::TNodeI NI) {
int j = 0;
int k = 0;
int prev = -1;
while (j < NI.GetInDeg() && k < NI.GetOutDeg()) {
int v1 = NI.GetInNId(j);
int v2 = NI.GetOutNId(k);
if (v1 <= v2) {
if (prev != v1) {
NeighbourV.Add(v1);
prev = v1;
}
j += 1;
} else {
if (prev != v2) {
NeighbourV.Add(v2);
prev = v2;
}
k += 1;
}
}
while (j < NI.GetInDeg()) {
int v = NI.GetInNId(j);
if (prev != v) {
NeighbourV.Add(v);
prev = v;
}
j += 1;
}
while (k < NI.GetOutDeg()) {
int v = NI.GetOutNId(k);
if (prev != v) {
NeighbourV.Add(v);
prev = v;
}
k += 1;
}
}
PJsonVal TGraphCascade::GetGraph() const {
PJsonVal G = TJsonVal::NewObj();
for (TNGraph::TNodeI NI = Graph.BegNI(); NI < Graph.EndNI(); NI++) {
TStr NodeNm = NodeIdNmH.GetDat(NI.GetId());
PJsonVal ParentsArr = TJsonVal::NewArr();
int InDeg = NI.GetInDeg();
for (int ParentN = 0; ParentN < InDeg; ParentN++) {
TStr ParentNm = NodeIdNmH.GetDat(NI.GetInNId(ParentN));
ParentsArr->AddToArr(ParentNm);
}
G->AddToObj(NodeNm, ParentsArr);
}
return G;
}
void plotParitialDegDistribution(const PNGraph& graph, std::vector<int>& nodeList) {
std::map<int, int> inDegDistMap;
std::map<int, int> outDegDistMap;
for (int i = 0; i < nodeList.size(); ++i) {
int curNodeId = nodeList[i];
if (!graph->IsNode(curNodeId)) continue;
TNGraph::TNodeI ni = graph->GetNI(curNodeId);
int curNodeInDeg = ni.GetInDeg();
if (inDegDistMap.find(curNodeInDeg) == inDegDistMap.end()) {
inDegDistMap.insert(std::pair<int, int>(curNodeInDeg, 0));
}
inDegDistMap[curNodeInDeg]++;
int curNodeOutDeg = ni.GetOutDeg();
if (outDegDistMap.find(curNodeOutDeg) == outDegDistMap.end()) {
outDegDistMap.insert(std::pair<int, int>(curNodeOutDeg, 0));
}
outDegDistMap[curNodeOutDeg]++;
}
TFltPrV inDegDist;
for (std::map<int, int>::iterator itr = inDegDistMap.begin(); itr != inDegDistMap.end(); itr++) {
inDegDist.Add(TFltPr(itr->first, itr->second));
}
TFltPrV outDegDist;
for (std::map<int, int>::iterator itr = outDegDistMap.begin(); itr != outDegDistMap.end(); itr++) {
outDegDist.Add(TFltPr(itr->first, itr->second));
}
TGnuPlot plot1("inDegDistParitial", "");
plot1.AddPlot(inDegDist, gpwPoints, "");
plot1.SetScale(gpsLog10XY);
plot1.SavePng();
TGnuPlot plot2("outDegDistParitial", "");
plot2.AddPlot(outDegDist, gpwPoints, "");
plot2.SetScale(gpsLog10XY);
plot2.SavePng();
TGnuPlot plot3("DegDistParitial", "");
plot3.AddCmd("set key right top");
plot3.AddPlot(inDegDist, gpwPoints, "In Degree");
plot3.AddPlot(outDegDist, gpwPoints, "Out Degree");
plot3.SetScale(gpsLog10XY);
plot3.SavePng();
}
void getInNeighborNodeIDs(const PNGraph& graph, int destNodeID, std::set<int>& nodeIdSet) {
std::queue<int> q;
q.push(destNodeID);
nodeIdSet.insert(destNodeID);
for (int level = 0; level < 2; ++level) {
int levelCount = q.size();
for (int i = 0; i < levelCount; ++i) {
int curNodeId = q.front();
q.pop();
// Scan neigbors;
TNGraph::TNodeI curNode = graph->GetNI(curNodeId);
int inDeg = curNode.GetInDeg();
for (int j = 0; j < inDeg; ++j) {
int curNeighborNodeID = curNode.GetInNId(j);
q.push(curNeighborNodeID);
nodeIdSet.insert(curNeighborNodeID);
}
}
}
}
double TCascade::GetProb(const PNGraph& G) {
double P = 0;
for (int n = 0; n < Len(); n++) {
const int DstNId = GetNode(n);
const double DstTm = GetTm(DstNId);
TNGraph::TNodeI NI = G->GetNI(DstNId);
double MxProb = log(Eps);
int BestParent = -1;
for (int e = 0; e < NI.GetInDeg(); e++) {
const int SrcNId = NI.GetInNId(e);
if (IsNode(SrcNId) && GetTm(SrcNId) < DstTm) {
const double Prob = log(TransProb(SrcNId, DstNId));
if (MxProb < Prob) { MxProb = Prob; BestParent = SrcNId; }
}
}
NIdHitH.GetDat(DstNId).Parent = BestParent;
P += MxProb;
}
return P;
}
void TSubGraphsEnum::RecurBfs(const int& NId, const int& Depth, TSimpleGraph& PrevG) {
if (Depth == 0) {
TIntPrV& EdgeV = PrevG();
EdgeV.Sort();
for (int i = 1; i < EdgeV.Len(); i++) {
if (EdgeV[i-1] == EdgeV[i]) { return; }
}
SgV.Add(PrevG);
return;
}
const TNGraph::TNodeI NI = NGraph ->GetNI(NId);
for (int e = 0; e < NI.GetOutDeg(); e++) {
TSimpleGraph CurG = PrevG;
CurG.AddEdge(NI.GetId(), NI.GetOutNId(e));
RecurBfs(NI.GetOutNId(e), Depth-1, CurG);
}
for (int e = 0; e < NI.GetInDeg(); e++) {
TSimpleGraph CurG = PrevG;
CurG.AddEdge(NI.GetInNId(e), NI.GetId());
RecurBfs(NI.GetInNId(e), Depth-1, CurG);
}
}
PNEANet KNNJaccard(PNGraph Graph, int K) {
PNEANet KNN = TNEANet::New();
int sum_neighbors = 0;
int ct;
int end;
end = Graph->GetNodes();
TIntV* Neighbors_old = new TIntV();
TIntV* Neighbors = new TIntV();
TIntV* temp;
TIntV NIdV;
Graph->GetNIdV (NIdV);
int size = NIdV.Len();
for (int ind = 0; ind < size; ind++) {
KNN->AddNode(NIdV[ind]);
}
KNN->AddFltAttrE("sim");
for (int ind = 0; ind < size; ind++) {
TNGraph::TNodeI NI = Graph->GetNI(NIdV[ind]);
if (NI.GetInDeg() > 0) {
continue;
}
if (NI.GetOutDeg() == 0) {
continue;
}
ct ++;
TVec<TPair<TFlt, TInt> > TopK;
for (int i = 0; i < K; i++) {
TopK.Add(TPair<TFlt,TInt>(0.0, -1));
}
Neighbors->Clr(false);
Neighbors_old->Clr(false);
for (int i = 0; i < NI.GetOutDeg(); i++) {
TNGraph::TNodeI Inst_NI = Graph->GetNI(NI.GetOutNId(i));
MergeNbrs(Neighbors, Neighbors_old, Inst_NI);
temp = Neighbors_old;
temp->Clr(false);
Neighbors_old = Neighbors;
Neighbors = temp;
}
int num = Neighbors_old->Len();
sum_neighbors += num;
//Swap neighbors and Neighbors_old
temp = Neighbors_old;
Neighbors_old = Neighbors;
Neighbors = temp;
for (int j = 0; j< Neighbors->Len(); j++) {
TNGraph::TNodeI Auth_NI = Graph->GetNI((*Neighbors)[j]);
float similarity = JaccardSim(NI, Auth_NI);
if (TopK[K-1].GetVal1() < similarity) {
int index = 0;
for (int i = K-2; i >= 0; i--)
if (TopK[i].GetVal1() < similarity) {
TopK.SetVal(i+1, TopK[i]);
} else {
index = i+1;
break;
}
TopK.SetVal(index, TPair<TFlt, TInt>(similarity, (*Neighbors)[j]));
}
}
for (int i = 0; i < K; i++) {
int EId = KNN->AddEdge(NI.GetId(), TopK[i].GetVal2());
KNN->AddFltAttrDatE(EId, TopK[i].GetVal1(), "sim");
}
// if (ct%10000 == 0)
// cout<<ct<<" avg neighbor degree = "<<sum_neighbors*1.0/ct<<" "<<currentDateTime()<<endl;
}
return KNN;
}
PNEANet KNNJaccardParallel(PNGraph Graph,int K) {
PNEANet KNN = TNEANet::New();
TIntV NIdV;
Graph->GetNIdV (NIdV);
int size = NIdV.Len();
for (int ind = 0; ind < size; ind++) {
KNN->AddNode(NIdV[ind]);
}
KNN->AddFltAttrE("sim");
TVec<TVec<TPair<TFlt, TInt>, int >, int > TopKList;
TVec<TVec<TPair<TFlt, TInt>, int >, int > ThTopK; // for each thread
TIntV NodeList;
TIntV ThNodeList;// for each thread
int NumThreads = omp_get_max_threads();
omp_set_num_threads(NumThreads);
#pragma omp parallel private(ThNodeList, ThTopK)
{
TIntV* Neighbors_old = new TIntV();
TIntV* Neighbors = new TIntV();
TIntV* temp;
#pragma omp for schedule(dynamic,1000)
for (int ind = 0; ind < size; ind++) {
TNGraph::TNodeI NI = Graph->GetNI(NIdV[ind]);
if (NI.GetInDeg() > 0) {
continue;
}
if (NI.GetOutDeg() == 0) {
continue;
}
TVec<TPair<TFlt, TInt>, int > TopK;
for (int i = 0; i < K; i++) {
TopK.Add(TPair<TFlt,TInt>(0.0, -1));
}
Neighbors->Clr(false);
Neighbors_old->Clr(false);
for (int i = 0; i < NI.GetOutDeg(); i++) {
TNGraph::TNodeI Inst_NI = Graph->GetNI(NI.GetOutNId(i));
MergeNbrs(Neighbors, Neighbors_old, Inst_NI);
temp = Neighbors_old;
temp->Clr(false);
Neighbors_old = Neighbors;
Neighbors = temp;
}
// Swap neighbors and Neighbors_old
temp = Neighbors_old;
Neighbors_old = Neighbors;
Neighbors = temp;
for(int j = 0; j< Neighbors->Len(); j++) {
TNGraph::TNodeI Auth_NI = Graph->GetNI((*Neighbors)[j]);
float similarity = JaccardSim(NI, Auth_NI);
if (TopK[K-1].GetVal1() < similarity) {
int index = 0;
for (int i = K-2; i >= 0; i--)
if (TopK[i].GetVal1() < similarity) {
TopK.SetVal(i+1, TopK[i]);
} else {
index = i+1;
break;
}
TopK.SetVal(index, TPair<TFlt, TInt>(similarity, (*Neighbors)[j]));
}
}
ThTopK.Add(TopK);
ThNodeList.Add(NIdV[ind]);
// if (ct%10000 == 0)
// cout<<ct<<" avg neighbor degree = "<<sum_neighbors*1.0/ct<<" "<<currentDateTime()<<endl;
}
#pragma omp critical
{
for (int j = 0; j < ThTopK.Len(); j++) {
TopKList.Add(ThTopK[j]);
NodeList.Add(ThNodeList[j]);
}
}
}
int size2 = NodeList.Len();
for (int i= 0; i < size2 ; i++) {
for (int j = 0; j < K; j++) {
if (TopKList[i][j].GetVal2() <= -1) {
break;
}
int EId = KNN->AddEdge(NodeList[i], TopKList[i][j].GetVal2());
KNN->AddFltAttrDatE(EId, TopKList[i][j].GetVal1(), "sim");
}
}
return KNN;
}
// Test node, edge creation
TEST(TNGraph, ManipulateNodesEdges) {
int NNodes = 10000;
int NEdges = 100000;
const char *FName = "test.graph.dat";
PNGraph Graph;
PNGraph Graph1;
PNGraph Graph2;
int i;
int n;
int NCount;
int x,y;
int Deg, InDeg, OutDeg;
Graph = TNGraph::New();
EXPECT_EQ(1,Graph->Empty());
// create the nodes
for (i = 0; i < NNodes; i++) {
Graph->AddNode(i);
}
EXPECT_EQ(0,Graph->Empty());
EXPECT_EQ(NNodes,Graph->GetNodes());
// create random edges
NCount = NEdges;
while (NCount > 0) {
x = (long) (drand48() * NNodes);
y = (long) (drand48() * NNodes);
// Graph->GetEdges() is not correct for the loops (x == y),
// skip the loops in this test
if (x != y && !Graph->IsEdge(x,y)) {
n = Graph->AddEdge(x, y);
NCount--;
}
}
EXPECT_EQ(NEdges,Graph->GetEdges());
EXPECT_EQ(0,Graph->Empty());
EXPECT_EQ(1,Graph->IsOk());
for (i = 0; i < NNodes; i++) {
EXPECT_EQ(1,Graph->IsNode(i));
}
EXPECT_EQ(0,Graph->IsNode(NNodes));
EXPECT_EQ(0,Graph->IsNode(NNodes+1));
EXPECT_EQ(0,Graph->IsNode(2*NNodes));
// nodes iterator
NCount = 0;
for (TNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
NCount++;
}
EXPECT_EQ(NNodes,NCount);
// edges per node iterator
NCount = 0;
for (TNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
for (int e = 0; e < NI.GetOutDeg(); e++) {
NCount++;
}
}
EXPECT_EQ(NEdges,NCount);
// edges iterator
NCount = 0;
for (TNGraph::TEdgeI EI = Graph->BegEI(); EI < Graph->EndEI(); EI++) {
NCount++;
}
EXPECT_EQ(NEdges,NCount);
// node degree
for (TNGraph::TNodeI NI = Graph->BegNI(); NI < Graph->EndNI(); NI++) {
Deg = NI.GetDeg();
InDeg = NI.GetInDeg();
OutDeg = NI.GetOutDeg();
EXPECT_EQ(Deg,InDeg+OutDeg);
}
// assignment
Graph1 = TNGraph::New();
*Graph1 = *Graph;
EXPECT_EQ(NNodes,Graph1->GetNodes());
EXPECT_EQ(NEdges,Graph1->GetEdges());
EXPECT_EQ(0,Graph1->Empty());
EXPECT_EQ(1,Graph1->IsOk());
// saving and loading
{
TFOut FOut(FName);
Graph->Save(FOut);
FOut.Flush();
}
{
TFIn FIn(FName);
Graph2 = TNGraph::Load(FIn);
}
EXPECT_EQ(NNodes,Graph2->GetNodes());
EXPECT_EQ(NEdges,Graph2->GetEdges());
EXPECT_EQ(0,Graph2->Empty());
EXPECT_EQ(1,Graph2->IsOk());
// remove all the nodes and edges
for (i = 0; i < NNodes; i++) {
n = Graph->GetRndNId();
Graph->DelNode(n);
}
EXPECT_EQ(0,Graph->GetNodes());
EXPECT_EQ(0,Graph->GetEdges());
EXPECT_EQ(1,Graph->IsOk());
EXPECT_EQ(1,Graph->Empty());
Graph1->Clr();
EXPECT_EQ(0,Graph1->GetNodes());
EXPECT_EQ(0,Graph1->GetEdges());
EXPECT_EQ(1,Graph1->IsOk());
EXPECT_EQ(1,Graph1->Empty());
}
// Node selects N~geometric(1.0-FwdBurnProb)-1 out-links and burns them. Then same for in-links.
// geometirc(p) has mean 1/(p), so for given FwdBurnProb, we burn 1/(1-FwdBurnProb)
void TForestFire::BurnGeoFire() {
const double OldFwdBurnProb = FwdBurnProb;
const double OldBckBurnProb = BckBurnProb;
const int& NInfect = InfectNIdV.Len();
const TNGraph& G = *Graph;
TIntH BurnedNIdH; // burned nodes
TIntV BurningNIdV = InfectNIdV; // currently burning nodes
TIntV NewBurnedNIdV; // nodes newly burned in current step
bool HasAliveInNbrs, HasAliveOutNbrs; // has unburned neighbors
TIntV AliveNIdV; // NIds of alive neighbors
int NBurned = NInfect, time;
for (int i = 0; i < InfectNIdV.Len(); i++) {
BurnedNIdH.AddDat(InfectNIdV[i]);
}
NBurnedTmV.Clr(false); NBurningTmV.Clr(false); NewBurnedTmV.Clr(false);
for (time = 0;; time++) {
NewBurnedNIdV.Clr(false);
for (int node = 0; node < BurningNIdV.Len(); node++) {
const int& BurningNId = BurningNIdV[node];
const TNGraph::TNodeI Node = G.GetNI(BurningNId);
// find unburned links
HasAliveOutNbrs = false;
AliveNIdV.Clr(false); // unburned links
for (int e = 0; e < Node.GetOutDeg(); e++) {
const int OutNId = Node.GetOutNId(e);
if (!BurnedNIdH.IsKey(OutNId)) {
HasAliveOutNbrs = true; AliveNIdV.Add(OutNId);
}
}
// number of links to burn (geometric coin). Can also burn 0 links
const int BurnNFwdLinks = Rnd.GetGeoDev(1.0 - FwdBurnProb) - 1;
if (HasAliveOutNbrs && BurnNFwdLinks > 0) {
AliveNIdV.Shuffle(Rnd);
for (int i = 0; i < TMath::Mn(BurnNFwdLinks, AliveNIdV.Len()); i++) {
BurnedNIdH.AddDat(AliveNIdV[i]);
NewBurnedNIdV.Add(AliveNIdV[i]); NBurned++;
}
}
// backward links
if (BckBurnProb > 0.0) {
// find unburned links
HasAliveInNbrs = false;
AliveNIdV.Clr(false);
for (int e = 0; e < Node.GetInDeg(); e++) {
const int InNId = Node.GetInNId(e);
if (!BurnedNIdH.IsKey(InNId)) {
HasAliveInNbrs = true; AliveNIdV.Add(InNId);
}
}
// number of links to burn (geometric coin). Can also burn 0 links
const int BurnNBckLinks = Rnd.GetGeoDev(1.0 - BckBurnProb) - 1;
if (HasAliveInNbrs && BurnNBckLinks > 0) {
AliveNIdV.Shuffle(Rnd);
for (int i = 0; i < TMath::Mn(BurnNBckLinks, AliveNIdV.Len()); i++) {
BurnedNIdH.AddDat(AliveNIdV[i]);
NewBurnedNIdV.Add(AliveNIdV[i]); NBurned++;
}
}
}
}
NBurnedTmV.Add(NBurned); NBurningTmV.Add(BurningNIdV.Len()); NewBurnedTmV.Add(NewBurnedNIdV.Len());
// BurningNIdV.AddV(NewBurnedNIdV); // node is burning eternally
BurningNIdV.Swap(NewBurnedNIdV); // node is burning just 1 time step
if (BurningNIdV.Empty()) break;
FwdBurnProb = FwdBurnProb * ProbDecay;
BckBurnProb = BckBurnProb * ProbDecay;
}
BurnedNIdV.Gen(BurnedNIdH.Len(), 0);
for (int i = 0; i < BurnedNIdH.Len(); i++) {
BurnedNIdV.Add(BurnedNIdH.GetKey(i));
}
FwdBurnProb = OldFwdBurnProb;
BckBurnProb = OldBckBurnProb;
}