Does Knowledge Distillation Really Work?
Samuel Don Stanton, Pavel Izmailov, Polina Kirichenko, Alexander A Alemi, Andrew Gordon Wilson
Abstract
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.
BibTeX
@inproceedings{
stanton2021does,
title={Does Knowledge Distillation Really Work?},
author={Samuel Don Stanton and Pavel Izmailov and Polina Kirichenko and Alexander A Alemi and Andrew Gordon Wilson},
booktitle={Advances in Neural Information Processing Systems},
editor={A. Beygelzimer and Y. Dauphin and P. Liang and J. Wortman Vaughan},
year={2021},
url={https://openreview.net/forum?id=7J-fKoXiReA}
}