Saved in:
Bibliographic Details
Main Authors: Pareek, Divyansh, Du, Simon S., Oh, Sewoong
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.04600
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910515182174208
author Pareek, Divyansh
Du, Simon S.
Oh, Sewoong
author_facet Pareek, Divyansh
Du, Simon S.
Oh, Sewoong
contents Self-Distillation is a special type of knowledge distillation where the student model has the same architecture as the teacher model. Despite using the same architecture and the same training data, self-distillation has been empirically observed to improve performance, especially when applied repeatedly. For such a process, there is a fundamental question of interest: How much gain is possible by applying multiple steps of self-distillation? To investigate this relative gain, we propose studying the simple but canonical task of linear regression. Our analysis shows that the excess risk achieved by multi-step self-distillation can significantly improve upon a single step of self-distillation, reducing the excess risk by a factor as large as $d$, where $d$ is the input dimension. Empirical results on regression tasks from the UCI repository show a reduction in the learnt model's risk (MSE) by up to 47%.
format Preprint
id arxiv_https___arxiv_org_abs_2407_04600
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Understanding the Gains from Repeated Self-Distillation
Pareek, Divyansh
Du, Simon S.
Oh, Sewoong
Machine Learning
Self-Distillation is a special type of knowledge distillation where the student model has the same architecture as the teacher model. Despite using the same architecture and the same training data, self-distillation has been empirically observed to improve performance, especially when applied repeatedly. For such a process, there is a fundamental question of interest: How much gain is possible by applying multiple steps of self-distillation? To investigate this relative gain, we propose studying the simple but canonical task of linear regression. Our analysis shows that the excess risk achieved by multi-step self-distillation can significantly improve upon a single step of self-distillation, reducing the excess risk by a factor as large as $d$, where $d$ is the input dimension. Empirical results on regression tasks from the UCI repository show a reduction in the learnt model's risk (MSE) by up to 47%.
title Understanding the Gains from Repeated Self-Distillation
topic Machine Learning
url https://arxiv.org/abs/2407.04600