R\'{e}nyi Divergence Deep Mutual Learning

TL;DR

This paper enhances Deep Mutual Learning by replacing KL divergence with Rényi divergence, offering better flexibility and improved generalization, supported by theoretical convergence analysis and extensive empirical validation.

Contribution

It introduces Rényi divergence into DML, providing a more flexible loss function and analyzing its convergence, which improves model performance and mitigates overfitting.

Findings

Rényi divergence improves DML performance.

Theoretical convergence with O(1) bias is established.

Empirical results show better generalization.

Abstract

This paper revisits Deep Mutual Learning (DML), a simple yet effective computing paradigm. We propose using R\'{e}nyi divergence instead of the KL divergence, which is more flexible and tunable, to improve vanilla DML. This modification is able to consistently improve performance over vanilla DML with limited additional complexity. The convergence properties of the proposed paradigm are analyzed theoretically, and Stochastic Gradient Descent with a constant learning rate is shown to converge with $\mathcal{O}(1)$-bias in the worst case scenario for nonconvex optimization tasks. That is, learning will reach nearby local optima but continue searching within a bounded scope, which may help mitigate overfitting. Finally, our extensive empirical results demonstrate the advantage of combining DML and R\'{e}nyi divergence, leading to further improvement in model generalization.