Keywords: distillation, deep ensembles, optimization
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.
Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.
TL;DR: We show that it is difficult to achieve high agreement between student and teacher in knowledge distillation and explain why
Supplementary Material: pdf
Code: https://github.com/samuelstanton/gnosis
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 2 code implementations](https://www.catalyzex.com/paper/does-knowledge-distillation-really-work/code)
17 Replies
Loading