Welcome to the Machine Learning for Fundamental Physics (ML4FP) School 2025, to be hosted at Lawrence Berkeley National Laboratory!
We invite particle physicists from across the field to come to register and spend a week at Berkeley learning about the wide range of machine learning methods that can be used to achieve the goals of particle physics. Though many examples from the program will be catered towards experimental applications, but the tools covered are general, so we welcome theorists to join us as well. We believe the best learning is done in action, so our program is structured to emphasized hands-on tutorial sessions rather than passive lectures.
Following the US ATLAS Machine Learning (ML) training events in 2022 and 2023, this programe evolved into the Machine Learning for Fundamental Physics School in 2024 with dedicated CMS and LZ sessions. This year we are excited to add a dedicated programming on neutrino physics! The school remains open to all particle physicists. Thanks to the financial assistance provided by US ATLAS through the ATC program, the accomodation for US ATLAS students at the LBL guesthouse can be covered by the school. There will also be a dedicated parallel sessions for ATLAS, CMS and neutrino physics on the final day.
All talks and tutorials will be given in person, and we encourage participants to come in person, although virtual participation is also welcome. For remote particpants, we can offer a limited number of GPU reservations on the NERSC system as well. However such a reservation is not strictly necessary to follow along with the program (particularly if one has access their own GPU resources), so optionally participants can register for remote participation without a GPU reservation.
Please note that all of the school's materials and recorded lectures will be made publically available, so anyone, regardless of registration, will be able to access it. We therefore encourage those registering for remote participation to only register if they plan on seriously following the program.
There is a 30 USD registration fees for this school for both in person participants and virtual participants using a GPU reservation. The total number of registrations for these participation modes will be capped, so please register soon. We will additionally allow remote registrations without a GPU reservation without a cap.
Program overview:
We will introduce fundamental concepts of machine learning accompanied by hands-on tutorials of the essential open-source ML packages. The program will cover diverse particle physics specific use cases with ample hands-on examples. There will be invited talks from veteran ML practitioners in HEP who have previously deployed ML solutons in their respective experiments.
Attendees can expect to gain an overview of the range of current and upcoming ML applications in HEP, and learn trade secrets on how to get ML to reliably work on particle physics datasets.
Tentative topics to be covered (many will include a hands-on component):
- Introduction to Machine Learning
- Introduction to standard open-source ML packages like Scikit-learn, Tensorflow and Pytorch
- Overview of ML in particle physics
- Overview of network architectures: symmetry preserving networks, transformers, and others
- Generative Models
- Anomaly Detection
- Neural Simulation-Based Inference
- Uncertainty Quantification
- Differentiable Programming
- Deploying NNs in real-time environments: Quantization, Model prunnning and compression, HLS4ML
The full agenda will be made public closer to the event.
Industry Talk:
In previous years we had speakers from Nvidia (Jaideep Pathak), Amazon (Chen Luo) and Google (Kanishka Rao) talk about their transition to industry and their research work in these companies. We will aim to have another excited industry talk this year.
Computing Resources:
In-person participants will be guaranteed computing resources for the hands-on sessions thanks to NERSC. Virtual participants that have registered for GPU resources will also receive NERSC training accounts. We may allow additional virtual particiants to audit the school without requesting GPU accounts, for whom there will be no registration fees.
Networking:
Past schools have led to new research collaobrations. This year's school will be another opportunity for young ML enthusiasts to connect with veteran ML experts in HEP.
Tutorial git:
Github link: https://github.com/ml4fp/2025-lbnl
Zoom:
Discussions:
Join the slack workspace to discuss and ask questions about the tutorials, particularly for remote participants.
Slack joining link:
Organizing team:
Aishik Ghosh, (UCI & LBL) [ATLAS and Neutrino Liason]
Elham E Khoda (SDSC/UCSD)
Benjamin Nachman (Stanford)
Sascha Diefenbacher (LBL) [ATLAS Liason]
Oz Amram (Fermilab) [CMS Liason]
Yifan Chen (SLAC) [Neutrino Liason]
Steering Committee:
Daniel Whiteson (UCI)
Shih-Chieh Hsu (UW)
Steven Farrell (NERSC/LBL)
Benjamin Nachman (Stanford)
Aishik Ghosh, (UCI & LBL)
Elham E Khoda (SDSC/UCSD)
