OL\'E: Orthogonal Low-rank Embedding, A Plug and Play Geometric Loss for Deep Learning

TL;DR

OLÉ introduces a geometric loss that enforces intra-class compactness and inter-class orthogonality in deep features, improving generalization and robustness without complex pair/triplet sampling.

Contribution

It proposes a novel plug-and-play loss that explicitly enforces intra-class similarity and inter-class margin through orthogonal low-rank subspace constraints, without requiring pair or triplet sampling.

Findings

Improves classification accuracy on object recognition tasks.

Enhances robustness and generalization of deep networks.

Achieves state-of-the-art results on Stanford STL-10.

Abstract

Deep neural networks trained using a softmax layer at the top and the cross-entropy loss are ubiquitous tools for image classification. Yet, this does not naturally enforce intra-class similarity nor inter-class margin of the learned deep representations. To simultaneously achieve these two goals, different solutions have been proposed in the literature, such as the pairwise or triplet losses. However, such solutions carry the extra task of selecting pairs or triplets, and the extra computational burden of computing and learning for many combinations of them. In this paper, we propose a plug-and-play loss term for deep networks that explicitly reduces intra-class variance and enforces inter-class margin simultaneously, in a simple and elegant geometric manner. For each class, the deep features are collapsed into a learned linear subspace, or union of them, and inter-class subspaces are pushed to be as orthogonal as possible. Our proposed Orthogonal Low-rank Embedding (OL\'E) does not require carefully crafting pairs or triplets of samples for training, and works standalone as a classification loss, being the first reported deep metric learning framework of its kind. Because of the improved margin between features of different classes, the resulting deep networks generalize better, are more discriminative, and more robust. We demonstrate improved classification performance in general object recognition, plugging the proposed loss term into existing off-the-shelf architectures. In particular, we show the advantage of the proposed loss in the small data/model scenario, and we significantly advance the state-of-the-art on the Stanford STL-10 benchmark.