We apologize again for inconvenience caused by MLcomp being down. Our
server (running on Amazon EC2) had a complete hardware failure, and we
had to spend some time recovering the data.
We took this opportunity to improve the site and are pleased to
announce three new features:
- A preliminary rating system for programs and datasets.
(Intuitively, good programs have low error rates and good datasets
distinguish programs - read more on http://mlcomp.org/faq).
- The dataset size limit has been increased from 50MB to 200MB (up to
1GB unzipped), as the backend has been improved.
- When you upload a program or dataset, new runs are automatically
created on it and the 5 highest rated dataset (or program).
Currently, this is done every 5 hours or so.
Please give us feedback on the new features!
-Percy and Jake