The Simpler The Better: An Entropy-Based Importance Metric To Reduce Neural Networks' Depth

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Quétu, Victor, Liao, Zhu, Tartaglione, Enzo
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913377484275712
author Quétu, Victor
Liao, Zhu
Tartaglione, Enzo
author_facet Quétu, Victor
Liao, Zhu
Tartaglione, Enzo
contents While deep neural networks are highly effective at solving complex tasks, large pre-trained models are commonly employed even to solve consistently simpler downstream tasks, which do not necessarily require a large model's complexity. Motivated by the awareness of the ever-growing AI environmental impact, we propose an efficiency strategy that leverages prior knowledge transferred by large models. Simple but effective, we propose a method relying on an Entropy-bASed Importance mEtRic (EASIER) to reduce the depth of over-parametrized deep neural networks, which alleviates their computational burden. We assess the effectiveness of our method on traditional image classification setups. Our code is available at https://github.com/VGCQ/EASIER.
format Preprint
id arxiv_https___arxiv_org_abs_2404_18949
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Simpler The Better: An Entropy-Based Importance Metric To Reduce Neural Networks' Depth
Quétu, Victor
Liao, Zhu
Tartaglione, Enzo
Machine Learning
While deep neural networks are highly effective at solving complex tasks, large pre-trained models are commonly employed even to solve consistently simpler downstream tasks, which do not necessarily require a large model's complexity. Motivated by the awareness of the ever-growing AI environmental impact, we propose an efficiency strategy that leverages prior knowledge transferred by large models. Simple but effective, we propose a method relying on an Entropy-bASed Importance mEtRic (EASIER) to reduce the depth of over-parametrized deep neural networks, which alleviates their computational burden. We assess the effectiveness of our method on traditional image classification setups. Our code is available at https://github.com/VGCQ/EASIER.
title The Simpler The Better: An Entropy-Based Importance Metric To Reduce Neural Networks' Depth
topic Machine Learning
url https://arxiv.org/abs/2404.18949