Distribution Shift Matters for Knowledge Distillation with Webly Collected Images

TL;DR

This paper introduces KD$^{3}$, a novel data-free knowledge distillation method that addresses distribution shift issues in webly collected data by instance selection, feature alignment, and distribution-invariant learning, improving performance.

Contribution

The paper proposes KD$^{3}$, a new approach that effectively handles distribution shifts in webly collected data for knowledge distillation, enhancing model reliability without original training data.

Findings

KD$^{3}$ outperforms existing data-free methods on benchmark datasets.

The method effectively mitigates distribution shift impacts.

Experimental results show improved student network performance.

Abstract

Knowledge distillation aims to learn a lightweight student network from a pre-trained teacher network. In practice, existing knowledge distillation methods are usually infeasible when the original training data is unavailable due to some privacy issues and data management considerations. Therefore, data-free knowledge distillation approaches proposed to collect training instances from the Internet. However, most of them have ignored the common distribution shift between the instances from original training data and webly collected data, affecting the reliability of the trained student network. To solve this problem, we propose a novel method dubbed ``Knowledge Distillation between Different Distributions" (KD$^{3}$), which consists of three components. Specifically, we first dynamically select useful training instances from the webly collected data according to the combined predictions of teacher network and student network. Subsequently, we align both the weighted features and classifier parameters of the two networks for knowledge memorization. Meanwhile, we also build a new contrastive learning block called MixDistribution to generate perturbed data with a new distribution for instance alignment, so that the student network can further learn a distribution-invariant representation. Intensive experiments on various benchmark datasets demonstrate that our proposed KD$^{3}$ can outperform the state-of-the-art data-free knowledge distillation approaches.