Hint-dynamic Knowledge Distillation

TL;DR

This paper introduces Hint-dynamic Knowledge Distillation (HKD), a novel method that adaptively utilizes teacher hints for each instance during training, improving the effectiveness of knowledge transfer in neural networks.

Contribution

HKD employs a meta-weight network and weight ensembling to dynamically and adaptively leverage teacher hints, enhancing knowledge distillation performance.

Findings

HKD outperforms existing methods on CIFAR-100 and Tiny-ImageNet.

Adaptive hint utilization improves student model accuracy.

Meta-weight and ensembling strategies effectively reduce bias.

Abstract

Knowledge Distillation (KD) transfers the knowledge from a high-capacity teacher model to promote a smaller student model. Existing efforts guide the distillation by matching their prediction logits, feature embedding, etc., while leaving how to efficiently utilize them in junction less explored. In this paper, we propose Hint-dynamic Knowledge Distillation, dubbed HKD, which excavates the knowledge from the teacher' s hints in a dynamic scheme. The guidance effect from the knowledge hints usually varies in different instances and learning stages, which motivates us to customize a specific hint-learning manner for each instance adaptively. Specifically, a meta-weight network is introduced to generate the instance-wise weight coefficients about knowledge hints in the perception of the dynamical learning progress of the student model. We further present a weight ensembling strategy to eliminate the potential bias of coefficient estimation by exploiting the historical statics. Experiments on standard benchmarks of CIFAR-100 and Tiny-ImageNet manifest that the proposed HKD well boost the effect of knowledge distillation tasks.