DeepJet Framework Swapneel Mehta, Mauro Verzetti, Jan Kieseler, - - PowerPoint PPT Presentation

▶

Mar 05, 2024 452 likes •788 views

DeepJet Framework Swapneel Mehta, Mauro Verzetti, Jan Kieseler, Markus Stoye, Maurizio Pierini CMS Experiment, EP-CMG-PS CERN Machine Learning 1. Comprehensive libraries 2. Fantastic documentation 3. Interactive Tutorials 4. Developer

SLIDE 1

DeepJet Framework

Swapneel Mehta, Mauro Verzetti, Jan Kieseler, Markus Stoye, Maurizio Pierini CMS Experiment, EP-CMG-PS CERN

SLIDE 2

SLIDE 3

SLIDE 4

SLIDE 5

SLIDE 6

SLIDE 7

Machine Learning

1. Comprehensive libraries 2. Fantastic documentation 3. Interactive Tutorials 4. Developer Community Support

SLIDE 8

Why build a library designed for high-energy physics?

SLIDE 9

Computer Scientists don’t always understand requirements for particle physics...

SLIDE 10

Physicists don’t always write great code...

SLIDE 11

Best of Both Worlds

1. Implement fast, efficient machine learning algorithms for physics 2. Provide high-level functions/wrappers for low-level tasks 3. Handle common bottlenecks - esp. memory -related issues 4. Create an extensible, easy-to-use framework

SLIDE 12

So what exactly is Jet Physics?

SLIDE 13

Jets: Collimated Streams of Particles

CMS Experiment

SLIDE 14

Michael Kagan

SLIDE 15

Michael Kagan

SLIDE 16

What does this library do?

SLIDE 17

Features of DeepJet

Data Conversion
Model Training
Prediction
Model Evaluation

SLIDE 18

Conversion

File-by-File
Avoids memory threshold

crossed (EOS)

Handles user-defined data

structures

Preprocessing support
Parallelized operation

SLIDE 19

Training

Keras-wrapped Tensorflow

backend

Additional callbacks
Monitor validity of tokens
Bookkeeping support

SLIDE 20

Prediction and Evaluation

Create compatible

prediction data structures

Support for Plots
Export of models and data

structures

SLIDE 21

Yeah, but why should I use it?

SLIDE 22

Simplicity

Modularised code, easy to

understand

Templates for quick-start
Step-by-step documentation
Elaborate examples and

use-cases

SLIDE 23

Support

Custom CPP Extensions

improve efficiency for Python

Automation of specific

tasks

Anaconda Environment

SLIDE 24

Upgrades

Available as a pip package

for Python 3.6

Tensorflow 1.8 supported
Integrating support for

TFRecords

Docker Image Distribution

SLIDE 25

Interesting! Tell me more about this library

SLIDE 26

DeepJetCore

SLIDE 27

DeepJet

SLIDE 28

DeepJet Demo

SLIDE 29

Conclusion

Easy-to-use Framework
Faster conversion and training
Diverse use-cases
Scalable to large datasets

SLIDE 30

Want to learn more about Machine Learning for High-energy Physics (MLHEP)?

SLIDE 31

Resources for Getting Started with MLHEP

https://github.com/iml-wg/HEP-ML-Resources https://www.coursera.org/learn/particle-physics [Shameless Plug] https://github.com/SwapneelM/awesome-particle- physics-for-non-physicists