Self-Distilled Self-Supervised Representation Learning
arXiv:2111.12958
Abstract
State-of-the-art frameworks in self-supervised learning have recently shown that fully utilizing transformer-based models can lead to performance boost compared to conventional CNN models. Striving to maximize the mutual information of two views of an image, existing works apply a contrastive loss to the final representations. Motivated by self-distillation in the supervised regime, we further exploit this by allowing the intermediate representations to learn from the final layer via the contrastive loss. Through self-distillation, the intermediate layers are better suited for instance discrimination, making the performance of an early-exited sub-network not much degraded from that of the full network. This renders the pretext task easier also for the final layer, leading to better representations. Our method, Self-Distilled Self-Supervised Learning (SDSSL), outperforms competitive baselines (SimCLR, BYOL and MoCo v3) using ViT on various tasks and datasets. In the linear evaluation and k-NN protocol, SDSSL not only leads to superior performance in the final layers, but also in most of the lower layers. Furthermore, qualitative and quantitative analyses show how representations are formed more effectively along the transformer layers. Code is available at https://github.com/hagiss/SDSSL.
WACV 23, 11 pages
References in corpus (10)
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
- Distilling the Knowledge in a Neural Network
- Learning Transferable Visual Models From Natural Language Supervision
- Bootstrap your own latent: A new approach to self-supervised Learning
- Learning Representations by Maximizing Mutual Information Across Views
- Deep Mutual Learning
- Representation Learning via Invariant Causal Mechanisms
- MultiGrain: a unified image embedding for classes and instances
- Decomposed Mutual Information Estimation for Contrastive Representation Learning