Data on the Brain & Mind

Overview

Our workshop aims to connect machine learning researchers with neuroscientists and cognitive scientists by focusing on concrete, open problems grounded in emerging neural datasets. Large-scale, open datasets are transforming the practice of research in a variety of neuroscience and cognitive science subdomains, including whole-brain circuitry (FlyWire; IBL; WormID), visual–motor cortical processing (MICrONS), human development (BabyView), and natural speech processing (MEG-MASC).

A key differentiator of our workshop is its focus on the rich diversity and heterogeneity of emerging neuroscience and cognitive science datasets. These are not your standard ML benchmarks; they span a wide array of data types including time series, images, videos, text, and even unconventional structures like connectome graphs from MICrONS. Each dataset possesses highly idiosyncratic meta-information, such as the spatial coordinates of electrodes in MEG recordings, that invites the development of tailored, sophisticated AI solutions. This departure from generic datasets encourages a move beyond one-size-fits-all foundation models, precisely the kinds of directions we aim to explore.

We believe a workshop focused on ML applied to such datasets is timely and presents a clear opportunity to help forge collaborations. Thus, we designed the workshop to be highly interactive. In addition, we invited speakers who have led large, open-sourced data initiatives or analyzed key datasets like those mentioned above.

There are three aims that shape the content and format of the workshop:

Aim 1: Identify open problems grounded in large, open-sourced datasets. What neuroscience and cognitive science questions can these large datasets address?
Aim 2: ML methods for neuroscience and cognitive science data. How can ML/AI methods help us collect datasets, analyze datasets, and, ultimately, answer scientific questions in a data-driven fashion?
Aim 3: Collaborative structures of ML applied to neuroscience and cognitive science. How can interdisciplinary collaborations between ML, neuroscience, and cognitive science researchers be structured to accelerate hypothesis generation and testing, particularly through shared tools for data-model integration and interpretability?

Event Schedule

Time	Event	Details/Speakers
7:30 - 8:00am	Poster Setup	Time for presenters to set up their posters.
8:00 - 8:10am	Opening Remarks
8:10 - 8:40am	Invited Talk 1	Dan Yamins (Stanford) Deep learning & visual cortex models
8:40 - 9:10am	Invited Talk 2	Ila Fiete (MIT) Neural computation, spatial navigation & memory
9:10 - 9:25am	Coffee break	Coffee break
9:25 - 9:55pm	Invited Talk 3	Laura Gwilliams (Stanford) Cognitive neuroscience of language processing
9:55 - 10:40pm	Contributed Talks 1	Lightning talks by authors of accepted papers.
10:40 - 11:30am	Poster Session 1	Interactive session with contributed works and discussion on datasets.
11:30am - 1:15pm	Lunch / Mentorship Lunch	Presenting authors of contributed works and invited speakers will participate in a mentorship lunch.
1:15 - 1:45pm	Invited Talk 4	Cengiz Pehlevan (Harvard) Models of neural computation & learning rules
1:45 - 2:15pm	Invited Talk 5	Rajesh Rao (University of Washington) Computational models of perception & decision-making
2:15 - 2:45pm	Contributed Talks 2	Lightning talks by authors of accepted papers.
2:45 - 3:15pm	Invited Talk 6	Bria Long (UCSD) Nature and development of visual concepts in human cognition + BabyView dataset
3:15 - 3:30pm	Coffee Break	Coffee break
3:30 - 4:10pm	Panel with invited speakers
4:10 - 4:15pm	Closing Remarks
4:15 - 5:00pm	Poster Session 2	Interactive session with contributed works and discussion on datasets.

Call for Papers

We invite submissions in the following two tracks. The review process is double-blind through our OpenReview portal for each track, and the submissions should be anonymized. Both tracks are considered non-archival will not conflict with future publication.

Reciprocal Review: To help ensure a fair and high-quality review process, we ask that each submission nominate at least one author to be available as a reviewer for the DBM workshop during September 9-19. It is sufficient if the nominated author has already signed up as a reviewer for the workshop.

Findings Track

Submissions should present original research on computational methods, theoretical frameworks, or quantitative findings that are applicable to neuroscience or cognitive science. We welcome 4-page or 8-page submissions of unpublished work (including work currently under review elsewhere). Previously published work is not eligible.

Examples:

Submission Template

Please use our official submission template when preparing your paper for the Findings track. Submissions not using the template may be desk-rejected. For authors new to LaTeX, we recommend using Overleaf. If you have any questions, please reach out to us.

Download the submission template (.zip)

Tutorials Track

Submissions should present pedagogical notebooks or blog posts discussing ML methods, neuroscience / cognitive science datasets, or conceptual frameworks, with the goal of facilitating interdisciplinary collaboration.

Examples:

Requirements

Submissions to the tutorials track will be evaluated on pedagogical quality and relevance to the themes of the workshop. Tutorials do not need to be original research—it is fine to present existing datasets or methods (with appropriate attribution) in a more accessible way. Any dependencies (code, data, etc.) should be included in the submission with clear instructions for use.

Additional instructions for submitting tutorials are provided here.

Frequently Asked Questions

Yes, please anonymize your submission.

We welcome concurrent submissions to both tracks, but each submission should stand on its own in content and contribution (and meet the standards of the track). Please see the calls for contributions for more details. We provide an option on OpenReview to indicate concurrent submission. In the case that the concurrent submissions of the same work are accepted to both tracks, the author(s) may consider presenting a single poster. We will help indicate that the poster presents both findings and a tutorial at the workshop.

The accepted papers for both tracks are considered non-archival and do not conflict with any future publications. The accepted papers will show up on OpenReview and can be cited from there. Work under review elsewhere is welcome.

For both tracks, all accepted submissions will be presented in the form of posters. Among the accepted submissions for both tracks, we will select and highlight the top submissions as oral talks in the workshop.

Yes, double submission is allowed as long as the work is on-topic for our workshop. At the same time, the goal of our workshop is to foster community participation and engagement. For that reason, we strongly encourage the presenting author to plan to attend the full workshop.