The correlation matrix in TMVA is -nan

[chaerme]

Hi,
I wrote a program similar to classification function in imitation of “TMVAClassification.C” and set three variables. However, when I tried to run my program, I found that the correlation matrix for each variable was’ -nan ', as follows.

<WARNING>    : Dataset[Default] : <GetCorrelationMatrix> Zero variances for variables (0, 1) = -nan
<WARNING>    : Dataset[Default] : <GetCorrelationMatrix> Zero variances for variables (0, 2) = -nan
<WARNING>   : Dataset[Default] : <GetCorrelationMatrix> Zero variances for variables (1, 0) = -nan
<WARNING>   : Dataset[Default] : <GetCorrelationMatrix> Zero variances for variables (1, 2) = -nan
<WARNING>  : Dataset[Default] : <GetCorrelationMatrix> Zero variances for variables (2, 0) = -nan
<WARNING>  : Dataset[Default] : <GetCorrelationMatrix> Zero variances for variables (2, 1) = -nan

As a result, the program cannot proceed smoothly.

<FATAL>   : <GetSeparation> signal and background histograms have different or invalid dimensions

Note that this program does read the data for each variable into TTree, because I can print it out with std ::cout (Or maybe I’m just too stupid to save the data into the Tree at all).
Perhaps there is a good person who can help answer this question?
Heartfelt thanks!
Kevin

[jalopezg]

Hi @chaerme; could you share the source code (and other related files) so that we can help you?

Cheers,
J.

[chaerme]

#include <cstdlib>
#include <iostream>
#include <fstream>
#include <vector>
#include <map>
#include <string>

#include "TChain.h"
#include "TFile.h"
#include "TTree.h"
#include "TString.h"
#include "TObjString.h"
#include "TSystem.h"
#include "TROOT.h"

#include "TMVA/Factory.h"
#include "TMVA/DataLoader.h"
#include "TMVA/Tools.h"
#include "TMVA/TMVAGui.h"

using namespace std;

//mylinux~> root -l TMVAtest.C\(\"myMethod1,myMethod2,myMethod3\"\)
int TMVAtest(TString myMethodList = "")
{
	// This loads the library
	TMVA::Tools::Instance();

	std::map<std::string, int> Use;

	
	Use["KNN"] = 1; // k-nearest neighbour method
	
	//Fisher
	Use["Fisher"] = 0;
	Use["FisherG"] = 0;
	Use["BoostedFisher"] = 0; // uses generalised MVA method boosting

	//ANN
	Use["MLP"] = 0; // Recommended ANN
	Use["MLPBFGS"] = 0; // Recommended ANN with optional training method
	Use["MLPBNN"] = 1; // Recommended ANN with BFGS training method and bayesian regulator
	Use["CFMlpANN"] = 0; // Depreciated ANN from ALEPH
	Use["TMlpANN"] = 0; // ROOT's own ANN

	 // Boosted Decision Trees
	Use["BDT"] = 1; // uses Adaptive Boost
	Use["BDTG"] = 0; // uses Gradient Boost
	Use["BDTB"] = 0; // uses Bagging
	Use["BDTD"] = 0; // decorrelation + Adaptive Boost
	Use["BDTF"] = 0; // allow usage of fisher discriminant for node splitting

	//**********HAVE NO RANDOM FOREST**************

	std::cout << std::endl;
	std::cout << "==> Start TMVAtest" << std::endl;

	// Select methods 
	if (myMethodList != "") {
		for (std::map<std::string, int>::iterator it = Use.begin(); it != Use.end(); it++) it->second = 0;

		std::vector<TString> mlist = TMVA::gTools().SplitString(myMethodList, ',');
		for (UInt_t i = 0; i < mlist.size(); i++) {
			std::string regMethod(mlist[i]);

			if (Use.find(regMethod) == Use.end()) {
				std::cout << "Method \"" << regMethod << "\" not known in TMVA under this name. Choose among the following:" << std::endl;
				for (std::map<std::string, int>::iterator it = Use.begin(); it != Use.end(); it++) std::cout << it->first << " ";
				std::cout << std::endl;
				return 1;
			}
			Use[regMethod] = 1;
		}
	}

	ifstream inputfile_s("./test_00.txt");
	if (!inputfile_s) {
		std::cout << "ERROR: could not open s_data file" << std::endl;
		exit(1);
	}

	ifstream inputfile_b("./test_01.txt");
	if (!inputfile_b) {
		std::cout << "ERROR: could not open p_data file" << std::endl;
		exit(1);
	}

	//Create Tree Model
	TTree* TreeS = new TTree("TreeS", "Tree for Signal and Background");
	Double_t var1, var2,var3;
	Double_t w;
	TreeS->Branch("var1", &var1, "var1");
	TreeS->Branch("var2", &var2, "var2");
	TreeS->Branch("var3", &var3, "var3");
	TreeS->Branch("w", &w, "w");

	TTree* TreeB = TreeS->CloneTree(0);

	//Read data
	vector <string> vec_s;
	vector <string> vec_b;
	vector <double> length_s;
	vector <double> length_b;
	vector <double> width_s;
	vector <double> width_b;
	vector <double> test_s;
	vector <double> test_b;
	string temp_s;
	string temp_b;

	while (getline(inputfile_s, temp_s))   //signal
	{
		vec_s.push_back(temp_s);
	}
	for (auto it = vec_s.begin(); it != vec_s.end(); it++)
	{
		istringstream is(*it);
		string s;
		int pam = 0;

		while (is >> s)
		{
			if (pam == 1)
			{
				double s_length = atof(s.c_str());
				length_s.push_back(s_length);
			}
			if (pam == 2)
			{
				double s_width;
				s_width = atof(s.c_str());
				width_s.push_back(s_width);
			}
			if (pam == 3)
			{
				double s_test;
				s_test = atof(s.c_str());
				test_s.push_back(s_test);
			}
			pam++;
		}
	}

	while (getline(inputfile_b, temp_b))   //background
	{
		vec_b.push_back(temp_b);
	}
	for (auto it = vec_b.begin(); it != vec_b.end(); it++)
	{
		istringstream is(*it);
		string b;
		int pam = 0;

		while (is >> b)
		{
			if (pam == 1)
			{
				double b_length = atof(b.c_str());
				length_b.push_back(b_length);
			}
			if (pam == 2)
			{
				double b_width;
				b_width = atof(b.c_str());
				width_b.push_back(b_width);
			}
			if (pam == 3)
			{
				double b_test;
				b_test = atof(b.c_str());
				test_b.push_back(b_test);
			}
			pam++;
		}
	}

	for (int i = 0; i < length_s.size(); i++)  //signal
	{
		w = 1.0;
		var1 = length_s.at(i);
		var2 = width_s.at(i);
		var3 = test_s.at(i);
		TreeS->Fill();
	}
	for (int i = 0; i < length_b.size(); i++)  //background
	{
		w = 1.0;
		var1 = length_s.at(i);
		var2 = width_s.at(i);
		var3 = test_b.at(i);
		TreeB->Fill();
	}

	//to save ntuples,histogram,etc.
	TString outfileName("TMVAoutput.root");
	TFile* outputFile = TFile::Open(outfileName, "RECREATE");

	//*********HAVETOCREATE*************
	TMVA::Factory* factory = new TMVA::Factory("TMVAtest", outputFile,
		"!V:!Silent:Color:DrawProgressBar:Transformations=I;D;P;G,D:AnalysisType=Classification");

	TMVA::DataLoader* dataloader = new TMVA::DataLoader("dataset");

	dataloader->AddVariable("var1", "Variable 1", "units", 'F');
	dataloader->AddVariable("var2", "Variable 2", "units", 'F');
	dataloader->AddVariable("var3", "Variable 3", "units", 'F');

	//there can add "spectator variables"

	dataloader->AddSignalTree(TreeS);
	dataloader->AddBackgroundTree(TreeB);
	dataloader->SetSignalWeightExpression("w");
	dataloader->SetBackgroundWeightExpression("w");

	//Apply additional cuts on the signal and background samples (can be different)
	TCut mycuts = ""; //for example :TCut mycuts = "abs(var1)<0.5 && abs(var2-0.5)<1";
	TCut mycutb = ""; //for example : TCut mycutb = "abs(var1)<0.5";

	//    dataloader->PrepareTrainingAndTestTree( mycut, "SplitMode=random:!V" );
	//    dataloader->PrepareTrainingAndTestTree( mycut,
   //         "NSigTrain=3000:NBkgTrain=3000:NSigTest=3000:NBkgTest=3000:SplitMode=Random:!V" );
	dataloader->PrepareTrainingAndTestTree(mycuts, mycutb, "nTrain_Signal=25:nTrain_Background=25:SplitMode=Random:NormMode=NumEvents:!V");

	//Book MVA Methods
	//KNN
	if (Use["KNN"])
		factory->BookMethod(dataloader, TMVA::Types::kKNN, "KNN",
			"H:nkNN=20:ScaleFrac=0.8:SigmaFact=1.0:Kernel=Gaus:UseKernel=F:UseWeight=T:!Trim");

	//Fisher
	if (Use["Fisher"])// Fisher discriminant (same as LD)
		factory->BookMethod(dataloader, TMVA::Types::kFisher, "Fisher", "H:!V:Fisher:VarTransform=None:CreateMVAPdfs:PDFInterpolMVAPdf=Spline2:NbinsMVAPdf=50:NsmoothMVAPdf=10");

	if (Use["FisherG"])// Fisher with Gauss-transformed input variables
		factory->BookMethod(dataloader, TMVA::Types::kFisher, "FisherG", "H:!V:VarTransform=Gauss");

	if (Use["BoostedFisher"])// Composite classifier: ensemble (tree) of boosted Fisher classifiers
		factory->BookMethod(dataloader, TMVA::Types::kFisher, "BoostedFisher",
			"H:!V:Boost_Num=20:Boost_Transform=log:Boost_Type=AdaBoost:Boost_AdaBoostBeta=0.2:!Boost_DetailedMonitoring");

	// Boosted Decision Trees
	if (Use["BDTG"]) // Gradient Boost
		factory->BookMethod(dataloader, TMVA::Types::kBDT, "BDTG",
			"!H:!V:NTrees=1000:MinNodeSize=2.5%:BoostType=Grad:Shrinkage=0.10:UseBaggedBoost:BaggedSampleFraction=0.5:nCuts=20:MaxDepth=2");

	if (Use["BDT"])  // Adaptive Boost
		factory->BookMethod(dataloader, TMVA::Types::kBDT, "BDT",
			"!H:!V:NTrees=850:MinNodeSize=2.5%:MaxDepth=3:BoostType=AdaBoost:AdaBoostBeta=0.5:UseBaggedBoost:BaggedSampleFraction=0.5:SeparationType=GiniIndex:nCuts=20");

	if (Use["BDTB"]) // Bagging
		factory->BookMethod(dataloader, TMVA::Types::kBDT, "BDTB",
			"!H:!V:NTrees=400:BoostType=Bagging:SeparationType=GiniIndex:nCuts=20");

	if (Use["BDTD"]) // Decorrelation + Adaptive Boost
		factory->BookMethod(dataloader, TMVA::Types::kBDT, "BDTD",
			"!H:!V:NTrees=400:MinNodeSize=5%:MaxDepth=3:BoostType=AdaBoost:SeparationType=GiniIndex:nCuts=20:VarTransform=Decorrelate");

	if (Use["BDTF"])  // Allow Using Fisher discriminant in node splitting for (strong) linearly correlated variables
		factory->BookMethod(dataloader, TMVA::Types::kBDT, "BDTF",
			"!H:!V:NTrees=50:MinNodeSize=2.5%:UseFisherCuts:MaxDepth=3:BoostType=AdaBoost:AdaBoostBeta=0.5:SeparationType=GiniIndex:nCuts=20");

	// Train MVAs using the set of training events
	factory->TrainAllMethods();

	// Evaluate all MVAs using the set of test events
	factory->TestAllMethods();

	// Evaluate and compare performance of all configured MVAs
	factory->EvaluateAllMethods();

	//Save the ouput
	outputFile->Close();

	std::cout << "==> Wrote root file: " << outputFile->GetName() << std::endl;
	std::cout << "==> TMVAtest is done!" << std::endl;

	delete factory;
	delete dataloader;
	// Launch the GUI for the root macros
	if (!gROOT->IsBatch()) TMVA::TMVAGui(outfileName);

	return 0;
}

int main(int argc, char** argv)
{
	// Select methods (don't look at this code - not of interest)
	TString methodList;
	for (int i = 1; i < argc; i++) {
		TString regMethod(argv[i]);
		if (regMethod == "-b" || regMethod == "--batch") continue;
		if (!methodList.IsNull()) methodList += TString(",");
		methodList += regMethod;
	}
	return TMVAtest(methodList);
}

Hi, @jalopezg ;
These are my source codes ,and I used Iris Flowers Data for the data file(ht( )tp://en.wikipedia.org/wiki/Iris_flower_data_set).I added the function of reading data in TXT files on the basis of TMVAClassification, and selected some methods I could use.
Thank you for your help!
Cheers,
Kevin

[jalopezg]

Hi @chaerme,

Could you attach here the test_00.txt and test_01.txt files for reproducibility? The problem might be more related to how the data is imported more than TMVA.

@chaerme

	while (getline(inputfile_s, temp_s))   //signal
	{
		vec_s.push_back(temp_s);
	}
	for (auto it = vec_s.begin(); it != vec_s.end(); it++)
	{
		istringstream is(*it);
		string s;
		int pam = 0;

Also, for creating a TTree out of a CSV file, see this related topic Read data from CSV file - #3 by renqipan. If your data consists of comma-separated values, this will both save some code and be less error prone.

Cheers,
J.

[chaerme]

Hi , @jalopezg
Since I am a newly registered user, I do not have the permission to upload files. I will try to copy the contents of TXT files here so that you can see them clearly：

Part of the content of test_00.txt       
1 5.1 3.5 1.4 0.2   I. setosa
2 4.9 3.0 1.4 0.2 I. setosa
3 4.7 3.2 1.3 0.2 I. setosa
4 4.6 3.1 1.5 0.2 I. setosa
5 5.0 3.6 1.4 0.3 I. setosa
6 5.4 3.9 1.7 0.4 I. setosa
7 4.6 3.4 1.4 0.3 I. setosa
8 5.0 3.4 1.5 0.2 I. setosa
9 4.4 2.9 1.4 0.2 I. setosa

Part of the content of test_01.txt
51 7.0 3.2 4.7 1.4 I. versicolor
52 6.4 3.2 4.5 1.5 I. versicolor
53 6.9 3.1 4.9 1.5 I. versicolor
54 5.5 2.3 4.0 1.3 I. versicolor
55 6.5 2.8 4.6 1.5 I. versicolor
56 5.7 2.8 4.5 1.3 I. versicolor
57 6.3 3.3 4.7 1.6 I. versicolor
58 4.9 2.4 3.3 1.0 I. versicolor
59 6.6 2.9 4.6 1.3 I. versicolor
60 5.2 2.7 3.9 1.4 I. versicolor
61 5.0 2.0 3.5 1.0 I. versicolor
62 5.9 3.0 4.2 1.5 I. versicolor
63 6.0 2.2 4.0 1.0 I. versicolor
64 6.1 2.9 4.7 1.4 I. versicolor
65 5.6 2.9 3.6 1.3 I. versicolor
66 6.7 3.1 4.4 1.4 I. versicolor
67 5.6 3.0 4.5 1.5 I. versicolor
68 5.8 2.7 4.1 1.0 I. versicolor

In the TXT file, I read columns 2 and 3 and 4 for each row. For example, using the first line of test_00.txt, 5.1, 3.5, 1.4 as var1, var2, and var3, respectively. The code I wrote above uses whitespace to determine which column the current data is in.

Cheers,
Kevin.

[jalopezg]

Hi @chaerme,

@chaerme

	var1 = length_s.at(i);
	var2 = width_s.at(i);
	var3 = test_b.at(i);

There is an obvious error here: length_s and width_s should be length_b and width_b instead. Can you try again after making these changes?

Cheers,
J.

[chaerme]

Hi, @jalopezg ,
It was such an embarrassing mistake that I didn’t notice it. However, when I corrected this error, the result was still the same, still desperate ‘-nan’.
I’ve just checked again and printed out a picture of the data using StartViewer, and I finally see the problem: TTree only stores the number of data, but not the value of each data. That is, it does access 50 instances, but the sample of each instance is’ 0 ‘, or ‘empty’, or something that’s not in the datafiles, so the correlation matrix is’ -nan '. But I have no idea how to solve this problem. I hope to get your help.

Cheers，
Kevin

[chaerme]

I think I’ve figured out the solution to the problem, although I haven’t figured out what ‘/D’ means yet (I’ll look it up later), but simply adding a ‘/D’ to each variable will store the value of the data in TTree and run TMVA smoothly.

**before**>>>>>>>>>>>TreeS->Branch("var1", &var1, "var1");

**now**>>>>>>>>>>>>TreeS->Branch("var1", &var1, "var1/D");

@jalopezg
Thanks a lots ! It was because of your help that I was able to find the problem step by step and get to the final solution. Thank you !

Cheers,
Kevin

[jalopezg]

Hi @chaerme,

@chaerme

I think I’ve figured out the solution to the problem, although I haven’t figured out what ‘/D’ means yet (I’ll look it up later), but simply adding a ‘/D’ to each variable will store the value of the data in TTree and run TMVA smoothly.

You can see that in the TTree documentation here. It tells TTree that the column type is a double-precision floating point number (Double_t).

@chaerme

Thanks a lots ! It was because of your help that I was able to find the problem step by step and get to the final solution. Thank you !

You are very welcome!

Cheers,
J.