AI competitions don’t produce useful models

Awesome post totally recommended: AI competitions don’t produce useful models

TLTR; From the upper average top performing models the one that has the best accuracy is not necessarily the best one. The blog post argues that the test set is not big enough to differentiate among the best models so in most of the cases what we find is which model actually better overfitted everything.

Leave a comment

Your email address will not be published. Required fields are marked *