2 Input MVA variables

Renata Kopecná edited this page 2022-01-31 16:10:46 +01:00

The variables used for the MVa training are stored in a .txt file with the name TMVA_variables_TheDecay_DL.txt, so in the case of the K+pi0 channel, the filename is TMVA_variables_KplusPi0Resolved.txt.

The file has to have a certain structure in order ot be read by the MVA_variables function.

The current file has the following structure

#ReaderName	 #LaTeXName 	#Unit 	
#DataType 		#NumberOfUsedBranches	#TFormula	#NamesOfBranches
gamma_max_log_PT:=max(log(gamma1_PT),log(gamma2_PT))	max(ln(p_{T}#gamma_{1}),ln(p_{T}#gamma_{2})	MeV	D	2	max(log(x),log(y))	gamma1_PT	gamma2_PT
K_plus_PI0_ETA:=TMath::Abs(pi_zero_resolved_ETA-K_plus_ETA)	|#eta(#pi^{0})-#eta(K^{+})|	0	D	2	abs(x-y)	pi_zero_resolved_ETA	K_plus_ETA
B_plus_NEW_ConePTasym	p_{T}ConeAsym(B^{+})	0	D	1	x	B_plus_NEW_ConePTasym
B_plus_log_PT:=log(B_plus_PT)	ln(p_{T}(B^{+}))	MeV	D	1	log(x)	B_plus_PT
B_plus_IPCHI2_OWNPV		B^{+}IP#chi^{2}			0	D	1	x	B_plus_IPCHI2_OWNPV
K_plus_log_PT:=log(K_plus_PT)	ln(p_{T}(K^{+}))	MeV	D	1	log(x)	K_plus_PT
B_plus_log_DIRA:=log(1-B_plus_DIRA_OWNPV) 	ln(1-B^{+}DIRA)		0	D	1	log(1-x)	B_plus_DIRA_OWNPV
pi_zero_resolved_CL	ConfidenceLevel#pi^{0}	0	D	1	x	pi_zero_resolved_CL
min_IPCHI2_mumu_OWNPV:=min(log(mu_plus_IPCHI2_OWNPV),log(mu_minus_IP_OWNPV))	min[ln(#mu^{+}IP#chi^{2})),ln(#mu^{-}IP#chi^{2}))]	0	D	2	min(log(x),log(y)) mu_plus_IPCHI2_OWNPV	mu_minus_IPCHI2_OWNPV
###

Each row represents one variable. The columns are separated by a tab. Each column represents some information:

1. The ROOT-readable variable name
2. The LaTeX name
3. The unit of the variable. If dimensionless, put 0
4. The data type. Either D (double), F (float) or I (integer). This is required to read the branches properly, ya know, ROOT....
5. How many branches are used for the variable? Typically 1 or 2
6. The TFormula used to calculate the variable. First branch is denoted by x, second by y. An example is max(log(x),log(y))
7. The names of the used branches

The last row is denoted by three crosses, ###. This signals to the reader it should stop reading. Anything below ### can be considered a comment.