Precedence for the separation of a classfier (<S^2>)?

Out of curiosity, I wanted to find more about the separation <S^2>, corresponding to equation (1) in section 3.1.10 in the TMVA User’s Guide. The reference [7] that the Guide links to appears to now be a dead link, so I haven’t been able to reference the thesis it appears to come from. So far I also haven’t been able to find elaboration about it in “The BABAR Physics Book”.

I would like to know if “separation” goes by a statistical distance name. A cursory look at Wikipedia leaves me wondering if it relates to “Energy distance”, and why it is used over “Hellinger distance”. The separation appears to return values closer to 1 than the Hellinger distance would.

Hi @jwruss,
sorry for the delay. @moneta might know, or alternatively you could try to contact the authors of the guide directly or hit the TMVA mattermost channel.


1 Like