Pruning By Explaining Revisited: Optimizing Attribution Methods to Prune CNNs and Transformers

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hatefi, Sayed Mohammad Vakilzadeh, Dreyer, Maximilian, Achtibat, Reduan, Wiegand, Thomas, Samek, Wojciech, Lapuschkin, Sebastian
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912082902908928
author Hatefi, Sayed Mohammad Vakilzadeh
Dreyer, Maximilian
Achtibat, Reduan
Wiegand, Thomas
Samek, Wojciech
Lapuschkin, Sebastian
author_facet Hatefi, Sayed Mohammad Vakilzadeh
Dreyer, Maximilian
Achtibat, Reduan
Wiegand, Thomas
Samek, Wojciech
Lapuschkin, Sebastian
contents To solve ever more complex problems, Deep Neural Networks are scaled to billions of parameters, leading to huge computational costs. An effective approach to reduce computational requirements and increase efficiency is to prune unnecessary components of these often over-parameterized networks. Previous work has shown that attribution methods from the field of eXplainable AI serve as effective means to extract and prune the least relevant network components in a few-shot fashion. We extend the current state by proposing to explicitly optimize hyperparameters of attribution methods for the task of pruning, and further include transformer-based networks in our analysis. Our approach yields higher model compression rates of large transformer- and convolutional architectures (VGG, ResNet, ViT) compared to previous works, while still attaining high performance on ImageNet classification tasks. Here, our experiments indicate that transformers have a higher degree of over-parameterization compared to convolutional neural networks. Code is available at https://github.com/erfanhatefi/Pruning-by-eXplaining-in-PyTorch.
format Preprint
id arxiv_https___arxiv_org_abs_2408_12568
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Pruning By Explaining Revisited: Optimizing Attribution Methods to Prune CNNs and Transformers
Hatefi, Sayed Mohammad Vakilzadeh
Dreyer, Maximilian
Achtibat, Reduan
Wiegand, Thomas
Samek, Wojciech
Lapuschkin, Sebastian
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
To solve ever more complex problems, Deep Neural Networks are scaled to billions of parameters, leading to huge computational costs. An effective approach to reduce computational requirements and increase efficiency is to prune unnecessary components of these often over-parameterized networks. Previous work has shown that attribution methods from the field of eXplainable AI serve as effective means to extract and prune the least relevant network components in a few-shot fashion. We extend the current state by proposing to explicitly optimize hyperparameters of attribution methods for the task of pruning, and further include transformer-based networks in our analysis. Our approach yields higher model compression rates of large transformer- and convolutional architectures (VGG, ResNet, ViT) compared to previous works, while still attaining high performance on ImageNet classification tasks. Here, our experiments indicate that transformers have a higher degree of over-parameterization compared to convolutional neural networks. Code is available at https://github.com/erfanhatefi/Pruning-by-eXplaining-in-PyTorch.
title Pruning By Explaining Revisited: Optimizing Attribution Methods to Prune CNNs and Transformers
topic Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2408.12568