Saved in:
Bibliographic Details
Main Authors: Hamidi, Shayan Mohajer, Deng, Xizhen, Tan, Renhao, Ye, Linfeng, Salamah, Ahmed Hussein
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.18041
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917733216550912
author Hamidi, Shayan Mohajer
Deng, Xizhen
Tan, Renhao
Ye, Linfeng
Salamah, Ahmed Hussein
author_facet Hamidi, Shayan Mohajer
Deng, Xizhen
Tan, Renhao
Ye, Linfeng
Salamah, Ahmed Hussein
contents Recently, it was shown that the role of the teacher in knowledge distillation (KD) is to provide the student with an estimate of the true Bayes conditional probability density (BCPD). Notably, the new findings propose that the student's error rate can be upper-bounded by the mean squared error (MSE) between the teacher's output and BCPD. Consequently, to enhance KD efficacy, the teacher should be trained such that its output is close to BCPD in MSE sense. This paper elucidates that training the teacher model with MSE loss equates to minimizing the MSE between its output and BCPD, aligning with its core responsibility of providing the student with a BCPD estimate closely resembling it in MSE terms. In this respect, through a comprehensive set of experiments, we demonstrate that substituting the conventional teacher trained with cross-entropy loss with one trained using MSE loss in state-of-the-art KD methods consistently boosts the student's accuracy, resulting in improvements of up to 2.6\%.
format Preprint
id arxiv_https___arxiv_org_abs_2407_18041
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle How to Train the Teacher Model for Effective Knowledge Distillation
Hamidi, Shayan Mohajer
Deng, Xizhen
Tan, Renhao
Ye, Linfeng
Salamah, Ahmed Hussein
Machine Learning
Recently, it was shown that the role of the teacher in knowledge distillation (KD) is to provide the student with an estimate of the true Bayes conditional probability density (BCPD). Notably, the new findings propose that the student's error rate can be upper-bounded by the mean squared error (MSE) between the teacher's output and BCPD. Consequently, to enhance KD efficacy, the teacher should be trained such that its output is close to BCPD in MSE sense. This paper elucidates that training the teacher model with MSE loss equates to minimizing the MSE between its output and BCPD, aligning with its core responsibility of providing the student with a BCPD estimate closely resembling it in MSE terms. In this respect, through a comprehensive set of experiments, we demonstrate that substituting the conventional teacher trained with cross-entropy loss with one trained using MSE loss in state-of-the-art KD methods consistently boosts the student's accuracy, resulting in improvements of up to 2.6\%.
title How to Train the Teacher Model for Effective Knowledge Distillation
topic Machine Learning
url https://arxiv.org/abs/2407.18041