X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=weighted.cpp;h=a0d593c803f691821afdb680e79be80f179b0afa;hb=5334a314651228c55fd0f206dec0209bc3148b24;hp=52d90c8ac4487d83acd1587b1d0f5c30b0c4cfe0;hpb=4745a956b3116a719f52f341d2a2db84df4817da;p=mothur.git diff --git a/weighted.cpp b/weighted.cpp index 52d90c8..a0d593c 100644 --- a/weighted.cpp +++ b/weighted.cpp @@ -22,8 +22,6 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { outputDir = o; numGroups = globaldata->Groups.size(); - - vector sums = getBranchLengthSums(t); if (m->control_pressed) { return data; } @@ -40,7 +38,7 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) if(processors == 1){ - data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums); + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size()); }else{ int numPairs = namesOfGroupCombos.size(); @@ -54,12 +52,12 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { lines.push_back(linePair(startPos, numPairsPerProcessor)); } - data = createProcesses(t, namesOfGroupCombos, sums); + data = createProcesses(t, namesOfGroupCombos); lines.clear(); } #else - data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums); + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size()); #endif return data; @@ -71,11 +69,10 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { } /**************************************************************************************************/ -EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos, vector& sums) { +EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos) { try { #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) int process = 1; - int num = 0; vector processIDS; EstOutput results; @@ -90,7 +87,7 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro }else if (pid == 0){ EstOutput Myresults; - Myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num, sums); + Myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num); m->mothurOut("Merging results."); m->mothurOutEndLine(); @@ -106,10 +103,14 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro out.close(); exit(0); - }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); } + }else { + m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine(); + for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); } + exit(0); + } } - results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num, sums); + results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num); //force parent to wait until all the processes are done for (int i=0;i<(processors-1);i++) { @@ -154,7 +155,7 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro } } /**************************************************************************************************/ -EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num, vector& sums) { +EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num) { try { EstOutput results; vector D; @@ -168,6 +169,7 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, string groupA = namesOfGroupCombos[h][0]; string groupB = namesOfGroupCombos[h][1]; + set validBranches; WScore[groupA+groupB] = 0.0; D.push_back(0.0000); //initialize a spot in D for each combination @@ -176,69 +178,75 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, map::iterator it = t->tree[t->groupNodeInfo[groupA][j]].pcount.find(groupA); int numSeqsInGroupI = it->second; - double weightedSum = ((numSeqsInGroupI * sums[t->groupNodeInfo[groupA][j]]) / (double)tmap->seqsPerGroup[groupA]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groupA][j], groupA, groupB); + double weightedSum = ((numSeqsInGroupI * sum) / (double)tmap->seqsPerGroup[groupA]); D[count] += weightedSum; } - + //adding the wieghted sums from group l for (int j = 0; j < t->groupNodeInfo[groupB].size(); j++) { //the leaf nodes that have seqs from group l map::iterator it = t->tree[t->groupNodeInfo[groupB][j]].pcount.find(groupB); int numSeqsInGroupL = it->second; - double weightedSum = ((numSeqsInGroupL * sums[t->groupNodeInfo[groupB][j]]) / (double)tmap->seqsPerGroup[groupB]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groupB][j], groupA, groupB); + double weightedSum = ((numSeqsInGroupL * sum) / (double)tmap->seqsPerGroup[groupB]); D[count] += weightedSum; } + count++; } //calculate u for the group comb - int total = t->getNumNodes(); - int twentyPercent = (total * 0.20); - for(int i=0;igetNumNodes();i++){ - for (int h = start; h < (start+num); h++) { + for (int h = start; h < (start+num); h++) { + //report progress + m->mothurOut("Processing combo: " + toString(h)); m->mothurOutEndLine(); + + string groupA = namesOfGroupCombos[h][0]; + string groupB = namesOfGroupCombos[h][1]; + + //calculate u for the group comb + for(int i=0;igetNumNodes();i++){ - string groupA = namesOfGroupCombos[h][0]; - string groupB = namesOfGroupCombos[h][1]; - - if (m->control_pressed) { return results; } + if (m->control_pressed) { return data; } double u; + //int pcountSize = 0; //does this node have descendants from groupA it = t->tree[i].pcount.find(groupA); //if it does u = # of its descendants with a certain group / total number in tree with a certain group if (it != t->tree[i].pcount.end()) { u = (double) t->tree[i].pcount[groupA] / (double) tmap->seqsPerGroup[groupA]; }else { u = 0.00; } - - + + //does this node have descendants from group l it = t->tree[i].pcount.find(groupB); //if it does subtract their percentage from u if (it != t->tree[i].pcount.end()) { u -= (double) t->tree[i].pcount[groupB] / (double) tmap->seqsPerGroup[groupB]; } - - u = abs(u * t->tree[i].getBranchLength()); - - //save groupcombs u value - WScore[(groupA+groupB)] += u; + + //if this is not the root then add it + if (rootForGrouping[namesOfGroupCombos[h]].count(i) == 0) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; + } + } + + } - //report progress - if((i) % twentyPercent == 0){ m->mothurOut("Percentage complete: " + toString(int((i / (float)total) * 100.0))); m->mothurOutEndLine(); } } - m->mothurOut("Percentage complete: 100"); m->mothurOutEndLine(); - /********************************************************/ //calculate weighted score for the group combination double UN; count = 0; for (int h = start; h < (start+num); h++) { UN = (WScore[namesOfGroupCombos[h][0]+namesOfGroupCombos[h][1]] / D[count]); - if (isnan(UN) || isinf(UN)) { UN = 0; } results.push_back(UN); count++; @@ -252,7 +260,7 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, } } /**************************************************************************************************/ -EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector& sums) { +EstOutput Weighted::getValues(Tree* t, string groupA, string groupB) { try { globaldata = GlobalData::getInstance(); @@ -263,6 +271,7 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector validBranches; vector groups; groups.push_back(groupA); groups.push_back(groupB); @@ -271,7 +280,8 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector::iterator it = t->tree[t->groupNodeInfo[groups[0]][j]].pcount.find(groups[0]); int numSeqsInGroupI = it->second; - double weightedSum = ((numSeqsInGroupI * sums[t->groupNodeInfo[groups[0]][j]]) / (double)tmap->seqsPerGroup[groups[0]]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groups[0]][j], groups[0], groups[1]); + double weightedSum = ((numSeqsInGroupI * sum) / (double)tmap->seqsPerGroup[groups[0]]); D += weightedSum; } @@ -281,40 +291,44 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector::iterator it = t->tree[t->groupNodeInfo[groups[1]][j]].pcount.find(groups[1]); int numSeqsInGroupL = it->second; - double weightedSum = ((numSeqsInGroupL * sums[t->groupNodeInfo[groups[1]][j]]) / (double)tmap->seqsPerGroup[groups[1]]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groups[1]][j], groups[0], groups[1]); + double weightedSum = ((numSeqsInGroupL * sum) / (double)tmap->seqsPerGroup[groups[1]]); D += weightedSum; } - + //calculate u for the group comb for(int i=0;igetNumNodes();i++){ - + if (m->control_pressed) { return data; } double u; + //int pcountSize = 0; //does this node have descendants from groupA it = t->tree[i].pcount.find(groupA); //if it does u = # of its descendants with a certain group / total number in tree with a certain group if (it != t->tree[i].pcount.end()) { u = (double) t->tree[i].pcount[groupA] / (double) tmap->seqsPerGroup[groupA]; }else { u = 0.00; } - - + + //does this node have descendants from group l it = t->tree[i].pcount.find(groupB); //if it does subtract their percentage from u if (it != t->tree[i].pcount.end()) { u -= (double) t->tree[i].pcount[groupB] / (double) tmap->seqsPerGroup[groupB]; } - - u = abs(u * t->tree[i].getBranchLength()); - - //save groupcombs u value - WScore[(groupA+groupB)] += u; - } - + + //if this is not the root then add it + if (rootForGrouping[groups].count(i) == 0) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; + } + } + } /********************************************************/ - + //calculate weighted score for the group combination double UN; UN = (WScore[(groupA+groupB)] / D); @@ -330,37 +344,71 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector Weighted::getBranchLengthSums(Tree* t) { +double Weighted::getLengthToRoot(Tree* t, int v, string groupA, string groupB) { try { - vector sums; + double sum = 0.0; + int index = v; + + //you are a leaf + if(t->tree[index].getBranchLength() != -1){ sum += abs(t->tree[index].getBranchLength()); } + double tempTotal = 0.0; + index = t->tree[index].getParent(); + + vector grouping; grouping.push_back(groupA); grouping.push_back(groupB); - for(int v=0;vgetNumLeaves();v++){ + rootForGrouping[grouping].insert(index); - if (m->control_pressed) { return sums; } + //while you aren't at root + while(t->tree[index].getParent() != -1){ + + if (m->control_pressed) { return sum; } - int index = v; - double sum = 0.0000; - - //while you aren't at root - while(t->tree[index].getParent() != -1){ - - //if you have a BL - if(t->tree[index].getBranchLength() != -1){ - sum += abs(t->tree[index].getBranchLength()); + //am I the root for this grouping? if so I want to stop "early" + //does my sibling have descendants from the users groups? + int parent = t->tree[index].getParent(); + int lc = t->tree[parent].getLChild(); + int rc = t->tree[parent].getRChild(); + + int sib = lc; + if (lc == index) { sib = rc; } + + map::iterator itGroup; + int pcountSize = 0; + itGroup = t->tree[sib].pcount.find(groupA); + if (itGroup != t->tree[sib].pcount.end()) { pcountSize++; } + itGroup = t->tree[sib].pcount.find(groupB); + if (itGroup != t->tree[sib].pcount.end()) { pcountSize++; } + + //if yes, I am not the root so add me + if (pcountSize != 0) { + if (t->tree[index].getBranchLength() != -1) { + sum += abs(t->tree[index].getBranchLength()) + tempTotal; + tempTotal = 0.0; + }else { + sum += tempTotal; + tempTotal = 0.0; + } + rootForGrouping[grouping].clear(); + rootForGrouping[grouping].insert(parent); + }else { //if no, I may be the root so add my br to tempTotal until I am proven innocent + if (t->tree[index].getBranchLength() != -1) { + tempTotal += abs(t->tree[index].getBranchLength()); } - index = t->tree[index].getParent(); - } - - //get last breanch length added - if(t->tree[index].getBranchLength() != -1){ - sum += abs(t->tree[index].getBranchLength()); } - sums.push_back(sum); + index = parent; } - return sums; + //get all nodes above the root to add so we don't add their u values above + index = *(rootForGrouping[grouping].begin()); + while(t->tree[index].getParent() != -1){ + int parent = t->tree[index].getParent(); + rootForGrouping[grouping].insert(parent); + index = parent; + } + + return sum; } catch(exception& e) { m->errorOut(e, "Weighted", "getBranchLengthSums"); @@ -370,4 +418,3 @@ vector Weighted::getBranchLengthSums(Tree* t) { /**************************************************************************************************/ -