Knowledge Distillation: A Survey

TL;DR

This survey reviews the field of knowledge distillation, a technique for compressing large neural networks into smaller, efficient models suitable for resource-constrained devices, highlighting methods, architectures, and future challenges.

Contribution

It provides a comprehensive overview of knowledge distillation, covering knowledge types, training schemes, architectures, algorithms, and applications, and discusses future research directions.

Findings

Knowledge distillation effectively creates small models from large ones.

Various distillation algorithms and architectures have been developed.

Challenges include improving performance and understanding theoretical foundations.

Abstract

In recent years, deep neural networks have been successful in both industry and academia, especially for computer vision tasks. The great success of deep learning is mainly due to its scalability to encode large-scale data and to maneuver billions of model parameters. However, it is a challenge to deploy these cumbersome deep models on devices with limited resources, e.g., mobile phones and embedded devices, not only because of the high computational complexity but also the large storage requirements. To this end, a variety of model compression and acceleration techniques have been developed. As a representative type of model compression and acceleration, knowledge distillation effectively learns a small student model from a large teacher model. It has received rapid increasing attention from the community. This paper provides a comprehensive survey of knowledge distillation from the perspectives of knowledge categories, training schemes, teacher-student architecture, distillation algorithms, performance comparison and applications. Furthermore, challenges in knowledge distillation are briefly reviewed and comments on future research are discussed and forwarded.