Source-linked AI summary
A Living Review of Machine Learning for Particle Physics
Matthew Feickert, Benjamin Nachman
TL;DR
Rapidly expanding ML research in HEP is difficult to track comprehensively. The paper presents a living, community-maintained review that organizes citations across HEP applications, with automated updating planned but necessarily incomplete.
Problem
Rapid development at the intersection of ML and HEP makes it difficult to follow current work and place new research in context.
Method
The paper describes a continuously updated, community-contributed citation review organized into manually assigned topics and sub-categories.
Results
The Living Review provides a nearly comprehensive citation list for ML papers applied or developed for experimental, phenomenological, or theoretical HEP analyses.
Takeaways & Limitations
The review is intended to help the HEP community keep track of rapid progress in machine learning research.
Takeaways & Limitations
The Living Review is not a general review of machine learning, so users should consult other sources for ML research outside HEP.
Abstract
from arXiv · showhide
Modern machine learning techniques, including deep learning, are rapidly being applied, adapted, and developed for high energy physics. Given the fast pace of this research, we have created a living review with the goal of providing a nearly comprehensive list of citations for those developing and applying these approaches to experimental, phenomenological, or theoretical analyses. As a living document, it will be updated as often as possible to incorporate the latest developments. A list of proper (unchanging) reviews can be found within. Papers are grouped into a small set of topics to be as useful as possible. Suggestions and contributions are most welcome, and we provide instructions for participating.
I. INTRODUCTION
The review addresses the difficulty of tracking rapidly developing ML-for-HEP research by maintaining a nearly comprehensive, continuously updated citation resource. It organizes relevant literature while distinguishing this living resource from general ML reviews and from peer-review validation.
- The review aims to provide a nearly comprehensive list of papers developing or applying ML to experimental, phenomenological, or theoretical HEP analyses.
- Rapid development at the intersection of ML and HEP makes recent work difficult for both new researchers and experienced practitioners to follow.
- The resource is continuously updated and open to community contributions to remain useful as the literature develops.
- Papers are organized into a small number of topics, and the Living Review also includes a list of unchanging reviews.
- Listing a paper does not endorse or validate its content, which remains a matter for the community and peer review.
- The paper introduces the Living Review's structure, contribution process, outlook, and conclusions, while serving as an unchanging reference for the review.
II. CATEGORIES
The review uses manually assigned, evolving categories and sub-categories to improve discoverability because paper metadata often does not support reliable automatic classification. Each category includes brief descriptions to help users navigate the resource.
- Manual topic organization improves discoverability because many papers lack keywords or sufficiently informative titles and abstracts for automatic categorization.
- Papers are manually placed into categories, may belong to multiple groups, and can be reclassified or expanded as the field evolves.
- Categories include Classification, Regression, Generation, and Anomaly Detection, with sub-categories for distinct directions within some topics.
- Brief descriptions are provided for each category and sub-category in the PDF form of the review.
- The Markdown website form hyperlinks topic papers to their references and, when available, their DOIs.
III. CONTRIBUTING
The Living Review accepts community contributions through GitHub pull requests and provides a contribution guide to support revisions and additions. The guide documents recommended procedures, software workflow, and common contributor questions.
- Community involvement complements the review's ongoing updates as new publications are released.
- Anyone may submit a new paper or document through a pull request to the review's GitHub project.
- The CONTRIBUTING.md guide gives detailed instructions and examples for revision and addition procedures and the associated software workflow.
IV. FUTURE PLANS
Future plans focus on automating reference updates and parts of the daily review update. The authors expect automation to remain incomplete because the field and its paper coverage change rapidly.
- The planned reference-update system would automatically update citations, potentially synchronizing preprints with INSPIRE through their links.
- Complete automation is considered unlikely because the field's rapidly changing nature affects what constitutes the field.
- Keyword queries may still identify a significant fraction of new papers posted to arXiv for daily updates.
- The proposed synchronization will not cover all papers because some are absent from arXiv or INSPIRE.
V. CONCLUSIONS
The paper presents the Living Review as an evolving resource for tracking machine-learning progress in high energy physics. It describes the review and encourages contributions to its continued development.
- The Living Review will continuously evolve as new machine-learning papers are written in high energy physics.
- The project welcomes community contributions to any aspect of the Living Review.
- The review is intended as a useful tool for keeping track of rapid progress in machine learning for high energy physics.