Kick-off meeting
Present:
Vincent, Mangi, Hieu, Balazs, Djalel, Roman, Naomi
Aim:
To develop and test ideas and concepts on how to improve the shower reconstruction and the separation of two showers in the Ecal and Hcal, using boosted decision trees for pattern recognition dof different shower types, with the end goal of improving Particle Flow Algorithms.
Naomi presented the analysis of hadrons in the Ecal prototype (work by Philippe Doublet) which is a starting point for further analysis. We discussed how boosted decision trees could be used in this context. A first step could be to improve the finding of the interaction layer, especially for "pointlike" events.
Balazs gave an introduction to machine learning and boosted decision trees.
The next meeting is planned for 20/11/12 at 9.30 am and will take place at LLR.
Mangi and Hieu will present PFA. We will discuss selection criteria for shower categories which could be used in MultiBoost.