English

Simple Unsupervised Knowledge Distillation With Space Similarity

Artificial Intelligence 2024-09-24 v1 Computer Vision and Pattern Recognition

Abstract

As per recent studies, Self-supervised learning (SSL) does not readily extend to smaller architectures. One direction to mitigate this shortcoming while simultaneously training a smaller network without labels is to adopt unsupervised knowledge distillation (UKD). Existing UKD approaches handcraft preservation worthy inter/intra sample relationships between the teacher and its student. However, this may overlook/ignore other key relationships present in the mapping of a teacher. In this paper, instead of heuristically constructing preservation worthy relationships between samples, we directly motivate the student to model the teacher's embedding manifold. If the mapped manifold is similar, all inter/intra sample relationships are indirectly conserved. We first demonstrate that prior methods cannot preserve teacher's latent manifold due to their sole reliance on L2L_2 normalised embedding features. Subsequently, we propose a simple objective to capture the lost information due to normalisation. Our proposed loss component, termed \textbf{space similarity}, motivates each dimension of a student's feature space to be similar to the corresponding dimension of its teacher. We perform extensive experiments demonstrating strong performance of our proposed approach on various benchmarks.

Keywords

Cite

@article{arxiv.2409.13939,
  title  = {Simple Unsupervised Knowledge Distillation With Space Similarity},
  author = {Aditya Singh and Haohan Wang},
  journal= {arXiv preprint arXiv:2409.13939},
  year   = {2024}
}
R2 v1 2026-06-28T18:52:03.895Z