Rethinking Generalizability and Discriminability of Self-Supervised Learning from Evolutionary Game Theory Perspective

TL;DR

This paper introduces a novel self-supervised learning approach inspired by evolutionary game theory, aiming to balance and improve both generalizability and discriminability of learned representations, supported by theoretical and empirical evidence.

Contribution

It proposes a new method combining EGT insights with reinforcement learning to optimize the trade-off between generalizability and discriminability in self-supervised learning.

Findings

Achieves state-of-the-art results on multiple benchmarks.

Theoretically tightens the generalization error bound.

Effectively balances representation properties during pre-training.

Abstract

Representations learned by self-supervised approaches are generally considered to possess sufficient generalizability and discriminability. However, we disclose a nontrivial mutual-exclusion relationship between these critical representation properties through an exploratory demonstration on self-supervised learning. State-of-the-art self-supervised methods tend to enhance either generalizability or discriminability but not both simultaneously. Thus, learning representations jointly possessing strong generalizability and discriminability presents a specific challenge for self-supervised learning. To this end, we revisit the learning paradigm of self-supervised learning from the perspective of evolutionary game theory (EGT) and outline the theoretical roadmap to achieve a desired trade-off between these representation properties. EGT performs well in analyzing the trade-off point in a two-player game by utilizing dynamic system modeling. However, the EGT analysis requires sufficient annotated data, which contradicts the principle of self-supervised learning, i.e., the EGT analysis cannot be conducted without the annotations of the specific target domain for self-supervised learning. Thus, to enhance the methodological generalization, we propose a novel self-supervised learning method that leverages advancements in reinforcement learning to jointly benefit from the general guidance of EGT and sequentially optimize the model to chase the consistent improvement of generalizability and discriminability for specific target domains during pre-training. Theoretically, we establish that the proposed method tightens the generalization error upper bound of self-supervised learning. Empirically, our method achieves state-of-the-art performance on various benchmarks.