First meeting of COMETA WG2: Technological innovation in data analysis

Europe/Zurich
Alessandra Cappati (Centre National de la Recherche Scientifique (FR)), Claudius Krause (Rutgers University), Riccardo Finotello (CEA Paris-Saclay)
Description

First meeting of COMETA WG2: Technological innovation in data analysis

 

WG2 Meeting

The WG2 (Technological innovation in data analysis) of COMETA (https://www.cost.eu/actions/CA22130/) met on 22nd November 2023 at 9 a.m. (to 10 a.m.) to officially start the discussions around data analysis in High Energy Physics (HEP) and Machine Learning (ML) techniques.

TL;DR

WHOWHATdetails
HEPIdentify suitable problemsThe HEP community should identify a list of problems solvable by ML techniques, possibly accompanied by methods to find data (datasets, simulations, etc.)
MLPropose techniques and benchmarksThe ML community should propose suitable techniques to address data analysis in HEP, and identify needs in terms of benchmarks/datasets
ALLContact peopleMore people in HEP+ML community should be involved in the action

Introduction

The main idea of the WG2 is to foster discussions around HEP problems which can be dealt with using ML. Viceversa, the ML community will welcome new benchmarks and datasets around which new techniques can be built and tested. On the one hand, advancements in ML techniques might lead to more advanced and precise data analysis. On the other hand, theoretical insight on ML algorithms might provide some new ways of exploring datasets [1].

Discussion

We started discussing on different HEP topics which might be interesting to start a discussion on ML methods:

  • jet/event topology tagging
  • polarizations
  • simulations
  • jet substructures
  • trajectory reconstruction
  • signal detection (to be tested on [2])

The existence of several datasets known to the HEP community was pointed out (e.g. [2] for continuous spectra detection and [3]). ML techniques could be directly tested on these datasets as an initial benchmark. Further discussions might then arise as to the informed use and development of these tools.

Future Works

We need to clearly identify a list of HEP issues that we would like to address using ML techniques. Some documents will be prepared to foster discussion and prepare future actions. On the ML side, we need to identify techniques and applications suitable for the tasks required in HEP.

More people involved in ML+HEP (e.g. [4]) should be involved in the action to share and propose new ideas. We should all see to that.


There are minutes attached to this event. Show them.