Spatial Entropy as an Inductive Bias for Vision Transformers

Elia Peruzzo, Enver Sangineto, Yahui Liu, Marco De Nadai, Wei Bi, Bruno Lepri, Nicu Sebe

Keywords: Vision Transformers, Self-Supervised Learning, Attention, Regularization.

TL;DR: We propose a self-supervised pretext task to include an object-based inductive bias in Vision Transformers.

Abstract: Recent work showed that the attention maps of Vision Transformers (VTs), when trained with self-supervision, can contain a semantic segmentation structure which does not spontaneously emerge when training is supervised. In this paper, we explicitly encourage the emergence of this spatial clustering as a form of training regularization, this way including a self-supervised pretext task into the standard supervised learning. In more detail, we exploit the assumption that, in a given image, objects usually correspond to few connected regions, and we propose a spatial formulation of the information entropy to quantify this object-based inductive bias. By minimizing the proposed spatial entropy, we include an additional self-supervised signal during training. Using extensive experiments, we show that the proposed regularization is beneficial with different training scenarios, datasets, downstream tasks and VT architectures. The code will be available upon acceptance.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Unsupervised and Self-supervised learning