Does Knowledge Distillation Really Work?

by   Samuel Stanton, et al.

Knowledge distillation is a popular technique for training a small student network to emulate a larger teacher model, such as an ensemble of networks. We show that while knowledge distillation can improve student generalization, it does not typically work as it is commonly understood: there often remains a surprisingly large discrepancy between the predictive distributions of the teacher and the student, even in cases when the student has the capacity to perfectly match the teacher. We identify difficulties in optimization as a key reason for why the student is unable to match the teacher. We also show how the details of the dataset used for distillation play a role in how closely the student matches the teacher – and that more closely matching the teacher paradoxically does not always lead to better student generalization.


Knowledge Distillation from Internal Representations

Knowledge distillation is typically conducted by training a small model ...

Learning Student-Friendly Teacher Networks for Knowledge Distillation

We propose a novel knowledge distillation approach to facilitate the tra...

What Knowledge Gets Distilled in Knowledge Distillation?

Knowledge distillation aims to transfer useful information from a teache...

Rethinking the Knowledge Distillation From the Perspective of Model Calibration

Recent years have witnessed dramatically improvements in the knowledge d...

Structural Knowledge Distillation

Knowledge distillation is a critical technique to transfer knowledge bet...

On the Efficacy of Knowledge Distillation

In this paper, we present a thorough evaluation of the efficacy of knowl...

Knowledge Distillation as Semiparametric Inference

A popular approach to model compression is to train an inexpensive stude...