Skip to Content.
Sympa Menu

fizinfo - [Fizinfo] RMI Elméleti Osztály Szemináriuma

fizinfo AT lists.kfki.hu

Subject: ELFT HÍRADÓ

List archive

[Fizinfo] RMI Elméleti Osztály Szemináriuma


Chronological Thread 
  • From: Balog Janos <balog.janos AT wigner.mta.hu>
  • To: <fizinfo AT lists.kfki.hu>, <rmkiusers AT lists.kfki.hu>
  • Subject: [Fizinfo] RMI Elméleti Osztály Szemináriuma
  • Date: Tue, 09 Dec 2014 08:50:55 +0100

RMI Elméleti Osztály Szemináriuma
Tisztelettel meghívjuk




Andrew Lowe
(Wigner ALICE Group)


"Quark/gluon jet tagging for ALICE: Machine learning for hadron physics using R"


címmel tartandó szemináriumára.


Kivonat:


Search strategies for new physics often depend on being able to efficiently discriminate between signal and background processes. To do this, we try to identify quantitative characteristics of each that will enable us to differentiate between them. In many cases, there are an enormous number of potential discriminant variables which we would like to combine to improve sensitivity. This naturally leads to the following question: how should we choose relevant discriminants that are minimally correlated with each other and provide the best predictive power? Similarly, given a newly-proposed discriminant variable, how can we rank this new variable against those we already know?

In this talk, I will outline systematic methods for ranking and selecting discriminants to maximise classification performance. Specifically, I will describe how these methods can be used to choose the best discriminant variables for use in a prototype jet tagging algorithm that could be used by the ALICE experiment at CERN for differentiating between quark-initiated and gluon-initiated jets. I will also introduce a new discriminant variables that have not previously been considered for quark/gluon jet tagging that were found using these methods.

This talk will present a walkthrough of the development of the prototype jet tagger and show promising preliminary results that are suggestive of the potential impact that this work could have on many physics analyses. Although presented in the context of jet tagging, the machine learning methods described in this talk could be of enormous benefit to the particle physics community and could lead to a marked improvement in discovery reach.


Helye: Wigner FK RMI III. ép. Tanácsterem
Ideje: 2014. december 11. csütörtök du. 14:00 óra




Szívesen látunk minden érdeklődőt.
Balog János





  • [Fizinfo] RMI Elméleti Osztály Szemináriuma, Balog Janos, 12/09/2014

Archive powered by MHonArc 2.6.19+.

Top of Page