JuliaHEP 2025 Workshop

ML for HEP Analysis in Julia

Not scheduled

25m

Princeton

Long Tutorial, 90-120' Talks

Speakers

Jerry 🦑 Ling (Harvard University (US))Dr Michael Savastio

Description

ML tutorial for JuliaHEP 2025 (90-120 minutes)

Part 1: Motivation and Basic analysis setup

We introduce basic data operation and JuliaHEP analysis setup by going through the follow topics:
- use Open Data
- implement cut and count
- simple visualization
- write to Arrow or Parquet prepare for ML

Part 2: Basic ML and intro to MLJ

We demonstrate how to conduct the most common types of HEP analysis ML tasks in Julia, incdluing:
- Using the simple target to train a two-class BDT as guide, introduce MLJ
- Basic data cleaning (sampling, balancing?)
- Show MLJ's capability of prototyping different ML models in quick succession

Part 3: AutoDiff, Lux, and more

Finally, we touch on some advanced topic such as automatic differentiate that are useful for certain tasks or suitable for building custom models using Lux.jl (such as auto encoder)