Back to $1M Lo-fi Mouse Emulation Prize

Open Call for Neural & Behavioral Datasets

Progress in computational neuroscience requires datasets that capture the full complexity of natural behavior, alongside corresponding neural activity. In this first stage of the prize, we are calling for proposals for generating neural and behavioral datasets of mice engaged in naturalistic, high-entropy behavior within a specific environment. While the primary goal of these datasets is to enable next-behavior prediction, they should also include neural recordings that could enhance the accuracy of such predictions. Selected datasets will be made public to enable the second stage of the prize, or used to evaluate models submitted in stage 2 and 3.

How to Participate

Submit a proposal for the dataset you want to generate in the form linked at the top of this page. Successful applicants will receive funding for developing their dataset. Submit your proposal as soon as possible – we are reviewing applications on a rolling basis, and will close this stage of the prize once we have enough datasets to move on to stage 2. We will award up to $300,000 in this stage. You can also apply to be awarded for existing, unpublished datasets. However, please note that our ability to award such datasets might depend on how they are licensed.

Judging criteria

Ambition: If successful, how well does the proposal contribute to the final goal of the prize?
Realism: Is this proposal cost-effective and possible to complete within a reasonable timeframe? We are aiming to complete this stage within around eight months, but the sooner the better.
Applicant strength: Can the applicant and/or the team pull off the proposal?

Bonus points if your proposal includes:

Naturalistic, freely moving behavior monitored exhaustively.
Long-term recordings.
Neural recordings with whole-brain coverage.
Recordings in multiple environments to measure out of distribution robustness of predictions.
Recordings in multiple labs to prevent overfitting to idiosyncracies of one particular environment.

Bonus questions you might address:

Existing datasets: please point out existing datasets/approaches we might have overlooked.
Scent, whisker touch, urine etc: while crucial, due to complexities in retrieving and integrating this data, it is currently out of scope for the proposal request. If you think differently and have ideas for how to generate and integrate such data, please let us know.
Out of distribution behavior: a stretch goal of the prize is to not only predict next behavior for a given environment (stage 2), but also out of distribution (OOD) behavior prediction (stage 3). If you have reason to believe that your proposed datasets and data generation approach is uniquely positioned to support the prediction of OOD later on – or you have any other ideas on how to handle OOD – please indicate so.
Tell us why we’re wrong: if you take issue with any of our assumptions and know a better way to achieve the long-term goal of the prize, tell us why.

Terms

Open sharing

Please note that, should you win the proposal, your datasets will be made public and contribute the basis for the second stage of the prize – creating models for next behavior prediction.

Funding

The winning proposals will be awarded a budget for generating the proposed datasets.

Eligibility

We welcome proposals from individuals, universities, research teams, companies or other institutions or formations that span or exceed these categories. We consider international proposals, apart from teams in US-sanctioned nations. If you apply for this first RFP stage of the prize, you may apply again for stage 2 and 3.

Gaming the Future: The Book!

Back to $1M Lo-fi Mouse Emulation Prize

Open Call for Neural & Behavioral Datasets

How to Participate

Judging criteria

Terms

Awarded Datasets

Search Foresight Institute