In-Person Poster presentation / poster accept
Long-Tailed Learning Requires Feature Learning
Thomas Laurent · James von Brecht · Xavier Bresson
MH1-2-3-4 #142
Keywords: [ long-tailed data distribution ] [ generalization ] [ deep learning theory ] [ Theory ]
We propose a simple data model inspired from natural data such as text or images, and use it to study the importance of learning features in order to achieve good generalization. Our data model follows a long-tailed distribution in the sense that some rare and uncommon subcategories have few representatives in the training set. In this context we provide evidence that a learner succeeds if and only if it identifies the correct features, and moreover derive non-asymptotic generalization error bounds that precisely quantify the penalty that one must pay for not learning features.