R\'{e}nyi Divergence Deep Mutual Learning
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 D...
Saved in:
Main Authors | , , , , , , |
---|---|
Format | Journal Article |
Language | English |
Published |
13.09.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | 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. |
---|---|
DOI: | 10.48550/arxiv.2209.05732 |