Science (Verification) Prize
- Repository with:
- Description of the problem
- The data
- Example code that reproduces the baseline
- Instructions for how to get the bot to run the code (part of the example code)
- Instructions for how the results should be output (so the bot can pick it up)
Have people:
-
fork it, modify the base with their model/parameters, add code in, etc
-
Create a PR when they are happy with the code + their local results,...
-
Bot will spin up AWS, run the code and grab the results, then display them as part of the PR
-
If a new PR comes in that beats an existing PR, it should comment/link to the old PR so that the user who was previously #1 is alerted about the new leader.
-
Add the result 'score' to some benchmark table in the example?
-
At the end of the competition, accept the PR with the best score?