Machine Learning and the Physical Sciences

About

Machine learning methods have had great success in learning complex representations of data that enable novel modeling and data processing approaches in many scientific disciplines. Physical sciences span problems and challenges at all scales in the universe: from finding exoplanets in trillions of sky pixels, to developing solutions to the quantum many-body problem and combinatorial problems, to detecting anomalies in event streams from the Large Hadron Collider, to predicting how extreme weather events will vary with climate change. Tackling a number of associated data-intensive tasks including, but not limited to, segmentation, computer vision, sequence modeling, causal reasoning, generative modeling, and probabilistic inference are critical for furthering scientific discovery in these and many other areas. In addition to using machine learning models for scientific discovery, the ability to interpret what a model has learned is receiving an increasing amount of attention.

In this targeted workshop, we aim to bring together computer scientists, mathematicians and physical scientists who are interested in applying machine learning to various outstanding physical problems including in inverse problems, approximating physical processes, understanding what a learned model represents, and connecting tools and insights from the physical sciences to the study of machine learning models. In particular, the workshop invites researchers to contribute short papers (extended abstracts) that demonstrate cutting-edge progress in the application of machine learning techniques to real-world problems in the physical sciences and/or using physical insights to understand and improve machine learning techniques.

By bringing together machine learning researchers and physical scientists who apply machine learning, we expect to strengthen the interdisciplinary dialogue, introduce exciting new open problems to the broader community, and stimulate the production of new approaches to solving challenging open problems in the sciences. Invited talks from leading individuals in both communities will cover the state-of-the-art techniques and set the stage for this workshop.

NeurIPS 2021

The Machine Learning and the Physical Sciences 2021 workshop will be held on December 13, 2021 as a part of the 35th Annual Conference on Neural Information Processing Systems. Originally planned to be at the Vancouver Convention Centre, Vancouver, BC, Canada, NeurIPS 2021 and this workshop will take place entirely virtually (online). Please check the main conference website for the latest information.

• 

  Official NeurIPS ML4PS schedule
  (for NeurIPS attendees)

• 

  Gather Town poster sessions
  Gather Town community development sessions
  (for NeurIPS attendees)

Schedule

Organizers

• 

  Atılım Güneş Baydin
  University of Oxford

• 

  Juan Felipe Carrasquilla
  Vector Institute / University of Waterloo

• 

  Adji Bousso Dieng
  Princeton University

• 

  Emine Kucukbenli
  Harvard University / Boston University

• 

  Gilles Louppe
  University of Liège

• 

  Benjamin Nachman
  Lawrence Berkeley National Laboratory

• 

  Brian Nord
  Fermilab

• 

  Savannah Thais
  Princeton University / IRIS-HEP

Steering Committee

• 

  Anima Anandkumar
  Caltech / NVIDIA

• 

  Kyle Cranmer
  New York University / Meta AI

• 

  Prabhat Ram
  Microsoft

• 

  Lenka Zdeborová
  EPFL

Papers/Posters

Call for papers

In this targeted workshop, we aim to bring together physical scientists and machine learning researchers who are interested in applying machine learning to various outstanding physical problems; or are using physical insights to understand and improve machine learning techniques.

We invite researchers to submit work particularly in the following and related areas:

• High-impact applications of machine learning to physical sciences, experiment or theory
• Strategies for incorporating prior scientific knowledge into machine learning algorithms
• Application of physical sciences to understand, model and improve machine learning techniques
• Machine learning model interpretability for obtaining insights to physical systems
• Any other area related to the subject of the workshop, including but not limited to techniques that are relevant to physical systems such as generative models, likelihood-free inference, variational inference, simulation-based inference, probabilistic models and so on.

Submissions of completed projects as well as high-quality works in progress are welcome. Submissions will be kept confidential until they are accepted and authors confirm that they can be included in the workshop. If a submission is not accepted, or withdrawn for any reason, it will be kept confidential and not made public.

Accepted work will be presented as posters during the workshop. Several accepted submissions will be selected for contributed talks. Please note that at least one coauthor of each accepted paper will be expected to have a NeurIPS conference registration that includes the workshop session and participate in one of the virtual poster sessions.

Important note for work that will be/has been published elsewhere:

All accepted short papers (extended abstracts) will be made available on the workshop website. This does not constitute an archival publication or formal proceedings; authors retain full copyright of their work and are free to publish their extended work in another journal or conference. We allow submission of extended abstracts that overlap with papers that are under review or have been recently published in a conference or a journal, including physical science journals. However, we do not accept cross submissions of the same content to multiple workshops at NeurIPS.

Submission instructions

Submissions should be anonymized short papers (extended abstracts) up to 4 pages in PDF format, typeset using the NeurIPS style. The authors are required to include a NeurIPS paper checklist, designed to encourage best practices for responsible machine learning research, addressing issues of reproducibility, transparency, research ethics, and societal impact. The paper checklist should come after the main paper content (see NeurIPS paper checklist guidelines). The paper checklist and references do not count towards the page limit. Appendices or supplementary material are discouraged, and reviewers are not expected to read beyond the first 4 pages and the impact statement. A workshop-specific modified NeurIPS style file will be provided for the camera-ready versions, after the author notification date.

Submissions page is here.

Submit paper

Instructions for Accepted Papers

Authors of accepted papers are expected to upload their camera-ready (final) paper and a poster by the deadlines given on this page. Optionally they can also record a short (5-minute) video describing their work.

Camera-ready papers

Please produce the "camera-ready" (final) version of your accepted paper by replacing the "neurips_2021.sty" style file with the "neurips_2021_ml4ps.sty" file available here and using the "final" package option (that is, "\usepackage[final]{neurips_2021_ml4ps}") to include author and affiliation information. The modified style file replaces the first page footer to correctly refer to the workshop instead of the main conference. It is acceptable if your paper goes up to five pages (excluding acknowledgments, references, paper checklist and any appendices if present) due to author and affiliation information taking extra space on the first page. The five-page limit is strict, and appendices are allowed but discouraged. Authors are encouraged, though not required, to include the NeurIPS Paper Checklist at the end of the paper (not counted in the page limit).

Please revise your paper as much as you can to reasonably address reviewer comments. The revision would include minor corrections and/or changes to directly address reviewer comments. Beyond these points, it is not acceptable to include any significant new material that was not present in the reviewed version of your paper.

Please upload the final PDF of your paper by the camera-ready deadline, by logging in to the CMT website (the same one used for the submissions) and using the camera-ready link shown with your existing submission.

Posters

The poster sessions will take place virtually in two GatherTown sessions. Posters will be presented during live and interactive sessions with virtual poster boards, where the presenter and the participants will be able to interact with audio and video. You will have control of your audio and video and can turn them on and off at any point as you wish. GatherTown is quite fun and it emulates a physical poster session venue where attendees can freely walk from poster to poster and interact in groups with the presenters through audio/video. Posters will be visible in full within the GatherTown platform and the attendees will have the possibility to zoom in to parts of your poster.

Please prepare an A0 landscape poster. The landscape orientation ensures that your poster is seen best in computer screens. We have observed a regular A0 format (as one would prepare for an in-person poster session where the posters are printed on A0 paper) also works fine in the GatherTown interface.

Please upload your poster by the poster upload deadline, using the central NeurIPS poster upload page and following the instructions given there regarding the file formats and resolutions.

In order to see the entry for your paper listed in the NeurIPS poster upload page, the co-author who is uploading the poster needs to be logged in to the neurips.cc website using the same email address they used in their paper submission. If you encounter a problem regarding NeurIPS accounts (e.g., you have multiple accounts associated with different email addresses and you need to merge these accounts into a single one), please consult the NeurIPS account FAQs and get in touch with the main NeurIPS conference organization who are handling accounts and registrations.

Optional videos

You can record a short video in addition to your poster, using the SlidesLive service contracted by NeurIPS to handle all video presentations. Uploaded videos will be available to watch on demand under the NeurIPS SlidesLive account. The video should be a brief (less than 5 minutes) presentation of your work in the accepted paper. Uploading a video is optional.

Authors of accepted papers will receive separate instructions from the SlidesLive team, who will share with you the URL and other info you need to use for video recording.

If you need some examples, you can see the posters and videos presented in the 2020 instance of the workshop here.

Important dates

• Submission deadline: September 18, September 27, 2021, 23:59 AoE
• Author notification: October 15, October 21, 2021, 23:59 PDT
• Camera-ready (final) paper deadline: November 26, 2021, 23:59 PDT
• Poster deadline: November 26, 2021, 23:59 PDT
• Workshop: December 13, 2021

Registration

NeurIPS conference has three main sessions (Tutorials, Conference, Workshops) to which you can register. You need to be registered to at least the Workshop session in order to be able to attend this workshop. For the latest registration-related information please refer to NeurIPS 2021 website.

Sponsors

• 
• 
• 

Location

Originally planned to be located in Sydney, Australia, NeurIPS 2021 and this workshop will take place entirely virtually (online). Please check the main conference website for the latest information.