DSD$^2$: Can We Dodge Sparse Double Descent and Compress the Neural Network Worry-Free?

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Quétu, Victor, Tartaglione, Enzo
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917584636477440
author Quétu, Victor
Tartaglione, Enzo
author_facet Quétu, Victor
Tartaglione, Enzo
contents Neoteric works have shown that modern deep learning models can exhibit a sparse double descent phenomenon. Indeed, as the sparsity of the model increases, the test performance first worsens since the model is overfitting the training data; then, the overfitting reduces, leading to an improvement in performance, and finally, the model begins to forget critical information, resulting in underfitting. Such a behavior prevents using traditional early stop criteria. In this work, we have three key contributions. First, we propose a learning framework that avoids such a phenomenon and improves generalization. Second, we introduce an entropy measure providing more insights into the insurgence of this phenomenon and enabling the use of traditional stop criteria. Third, we provide a comprehensive quantitative analysis of contingent factors such as re-initialization methods, model width and depth, and dataset noise. The contributions are supported by empirical evidence in typical setups. Our code is available at https://github.com/VGCQ/DSD2.
format Preprint
id arxiv_https___arxiv_org_abs_2303_01213
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle DSD$^2$: Can We Dodge Sparse Double Descent and Compress the Neural Network Worry-Free?
Quétu, Victor
Tartaglione, Enzo
Machine Learning
Neoteric works have shown that modern deep learning models can exhibit a sparse double descent phenomenon. Indeed, as the sparsity of the model increases, the test performance first worsens since the model is overfitting the training data; then, the overfitting reduces, leading to an improvement in performance, and finally, the model begins to forget critical information, resulting in underfitting. Such a behavior prevents using traditional early stop criteria. In this work, we have three key contributions. First, we propose a learning framework that avoids such a phenomenon and improves generalization. Second, we introduce an entropy measure providing more insights into the insurgence of this phenomenon and enabling the use of traditional stop criteria. Third, we provide a comprehensive quantitative analysis of contingent factors such as re-initialization methods, model width and depth, and dataset noise. The contributions are supported by empirical evidence in typical setups. Our code is available at https://github.com/VGCQ/DSD2.
title DSD$^2$: Can We Dodge Sparse Double Descent and Compress the Neural Network Worry-Free?
topic Machine Learning
url https://arxiv.org/abs/2303.01213