About the challenge
We are happy to announce the first Seldonian Toolkit Competition, a competition that allows undergraduate and graduate students in the US and Canada to learn about and develop safe and fair machine learning algorithms using the recently launched Seldonian Toolkit. This toolkit was created to make it easier for data scientists to apply machine learning responsibly: with high-confidence safety and fairness constraints on the machine learning system's behavior.
We welcome people from different fields (not just computer science!) to participate in this competition. Students will have the freedom to select the application and dataset that they work with, allowing them to tackle problems within their own discipline using our machine learning tools and with expert support.
Time Commitment
The open-ended nature of this competition means that participants could spend a relatively small amount of time, like a weekend, putting together a submission. However, participants might also devote significant time to their project over the 1-2 months that the contest runs. Overall, we anticipate that the most successful teams will spend at least 3-5 hours per week on the project for the 1-2 months that the contest runs.
Requirements
Project Topic
This competition is intentionally open ended. This means that instead of us picking a specific dataset or problem for you to solve, you get to pick a topic that you find interesting! For some examples of possible projects, we recommend reviewing the various tutorials and examples that can be found using the links at the top of this page. These include a variety of examples where fairness with respect to properties like age, race, religion, and gender is important. One can also enforce safety constraints by placing accuracy requirements on models, or by defining additional safety related constraints (see for example the safety constraints enforced on a machine learning model for type 1 diabetes treatment, as described in this paper). Though teams can use the toolkit for supervised learning (regression and classification) applications and reinforcement learning applications, we expect that most applications will focus on supervised learning settings, as the reinforcement learning components of the toolkit are still relatively nascent.
What to Submit
Prizes
Best Overall Project
Projects will be evaluated based on the potential positive impacts of the proposed application, the performance of the trained system, the feasibility of the proposed application to the real-world, and the clarity and quality of the submitted report and code. All teams are eligible for this award.
Devpost Achievements
Submitting to this hackathon could earn you:
Judges
Phillp Thomas
Associate Professor/UMass Amherst
Judging Criteria
-
TODO
TODO
Questions? Email the hackathon manager
Tell your friends
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.