Improve Knowledge Distillation via Label Revision and Data Selection

TL;DR

This paper introduces a novel knowledge distillation method that improves student model training by revising teacher predictions and selecting appropriate data samples, addressing supervision reliability issues.

Contribution

It proposes Label Revision and Data Selection techniques to enhance knowledge distillation by reducing supervision errors and improving performance.

Findings

Method improves student accuracy over vanilla KD.

Combines effectively with existing distillation approaches.

Reduces impact of erroneous teacher predictions.

Abstract

Knowledge distillation (KD) has become a widely used technique in the field of model compression, which aims to transfer knowledge from a large teacher model to a lightweight student model for efficient network development. In addition to the supervision of ground truth, the vanilla KD method regards the predictions of the teacher as soft labels to supervise the training of the student model. Based on vanilla KD, various approaches have been developed to further improve the performance of the student model. However, few of these previous methods have considered the reliability of the supervision from teacher models. Supervision from erroneous predictions may mislead the training of the student model. This paper therefore proposes to tackle this problem from two aspects: Label Revision to rectify the incorrect supervision and Data Selection to select appropriate samples for distillation to reduce the impact of erroneous supervision. In the former, we propose to rectify the teacher's inaccurate predictions using the ground truth. In the latter, we introduce a data selection technique to choose suitable training samples to be supervised by the teacher, thereby reducing the impact of incorrect predictions to some extent. Experiment results demonstrate the effectiveness of our proposed method, and show that our method can be combined with other distillation approaches, improving their performance.