Could learning machines keep getting better?
We would like to gradually move towards learning machines that are fully automatic. This can only happen if they get exposed to more and more data. The Life Long Machine Learning challenge in machine learning ambitions to keep stimulating the community towards ever pushing the frontiers of Automatic Machine Learning.
We envision that the last phase of the AutoML challenge will give us a starting point with:
- 5 reference AutoML packages.
- 5 reference datasets.
From there, people will be able to
- propose new software packages to beat the current 5 reference AutoML machines on the reference datasets, and
- submit new datasets to give harder benchmarks for the current 5 reference AutoML machines.
Every 6 months, we will change the AutoML reference machine and the datasets.
Join us if you want to contribute or have better ideas, post a message on the forum!