AI4HallC Working Group Meeting

Friday Feb 9, 2024, 3:00 PM → 5:20 PM US/Eastern

Casey Morean (Catholic University of America), Cristiano Fanelli (William & Mary, Jefferson Lab), Mark Jones (Jefferson Lab), Tanja Horn (Catholic University of America)

AI/ML has become ubiquitous in nuclear physics in the last few years and new possibilities have been emerging. This is an opportune time for Hall C to take advantage of these developments in computing technologies and statistical methods and define its path forward. The main goal of the Hall C AI/ML Working Group will be to provide a forum for discussion for anyone interested in defining this path, exploring possible applications of AI/ML in Hall C, and connecting to data scientists.

 

Hall C is the precision measurements hall at JLab providing pillars of measurements to constrain physics quantities like Parton Distribution Function and Parton Distribution Amplitudes. Because of this Hall C has a unique role and very different equipment and operation requirements from the other halls, e.g., Hall C operation has multiple subsystems (beam, spectrometer, target) whose drifts/changes must be monitored as they all directly impact performance and physics output of Hall C. 

 

We had two previous productive meetings where two major needs were identified and discussed:

1.) Operations composed of optics, data preparation, and equipment and operation

2.) Uncertainty quantification for global physics analysis, e.g., PDF/PDA

For operations, a list of parameters relevant for monitoring was determined.  A need for hands-on examples was also identified.  More details are available in the previous live notes and parameter list.

 

At this meeting, several ongoing efforts from the community will be presented. Members of the Hall C community have applied computer vision techniques, LLMs, and unsupervised learning to Hall C data.  The implementation of Hydra in Hall C will be discussed.  Efforts from the broader lab community / DOE will be shown. We will continue the discussion on operations and uncertainty quantification before moving to a Hall C specific hands-on example.

 

Zoom Link

Recording (TBD)

 

Google Live Document Link: 

https://docs.google.com/document/d/1dfQ4-cLT8QWV3-_C2RT5YlXQOBx-E-x98C5qNMlkmGc/edit?usp=sharing

Introduction

Live Demonstration

In this demonstration, we will leverage the Jupyter notebook server at JLab. Example yet to be determined.

Convener: Casey Morean (Catholic University of America)

Presentations from the Community

Applications of AI/ML towards data from Hall C. In this session, Carlos Yero will present work done in interpreting angle information using computer vision. Cameron Cotton will present his work on PID in the HMS using unsupervised learning. Casey Morean will present a toy implementation of retrieval augmented generation (RAG) from the Hall C logbook and wiki.

Overview of broader AI/ML program at JLab

Discussion