A question about how to set the classifier range

TMVA
1

Dear experts,

When I booked the Fisher method and get the output classifier, I will always got the classifier in range[-0.6,0.6]. While in our analysis we typically want a classifier in range [0,1].

Although I know I can do a linear transformation with the output classifier from the range [-0.6,0.6] to range [0,1] manually, by myself, to get the expected range we want, I still wonder if there is any possibility to set this range automatically before we get the classifier? (So we don’t need to do other transformation to get the range of out classifier we expected) I failed to find the solution by reading the documentation. :sob:

Best regards.

截屏2022-11-21 14.37.31
截屏2022-11-21 14.37.311332×974 110 KB

2

I guess @moneta can help with this

3

Hi,
If you want the classifier to return an output in the range [0,1] , i.e. representing a probability, you can use the classifier probability distributions which gives the probability for the event to be of signal type or the rarity which is the cumulative integral of that probability. Both are in the range between [0,1].
See page 29 of the TMVA Users Guide

Best regards

Lorenzo

4

Thanks @moneta , sorry for the delaying response, this reply answered my question perfectly!