Self-supervised learning disentangled group representation as feature

• Main Authors: WANG, Tan, YUE, Zhongqi, HUANG, Jianqiang, SUN, Qianru, ZHANG, Hanwang
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2021
• Subjects: Databases and Information Systems; Graphics and Human Computer Interfaces
• Online Access: https://ink.library.smu.edu.sg/sis_research/6227; https://ink.library.smu.edu.sg/context/sis_research/article/7230/viewcontent/NIPS2021_Causal_SSL_final.pdf
• Institution: Singapore Management University

A good visual representation is an inference map from observations (images) to features (vectors) that faithfully reflects the hidden modularized generative factors (semantics). In this paper, we formulate the notion of “good” representation from a group-theoretic view using Higgins’ definition of disentangled representation [38], and show that existing Self-Supervised Learning (SSL) only disentangles simple augmentation features such as rotation and colorization, thus unable to modularize the remaining semantics. To break the limitation, we propose an iterative SSL algorithm: Iterative Partition-based Invariant Risk Minimization (IP-IRM), which successfully grounds the abstract semantics and the group acting on them into concrete contrastive learning. At each iteration, IP-IRM first partitions the training samples into two subsets that correspond to an entangled group element. Then, it minimizes a subset-invariant contrastive loss, where the invariance guarantees to disentangle the group element. We prove that IP-IRM converges to a fully disentangled representation and show its effectiveness on various benchmarks.