Data-Free Dynamic Compression of CNNs for Tractable Efficiency

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Meiner, Lukas, Mehnert, Jens, Condurache, Alexandru Paul
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909542881689600
author Meiner, Lukas
Mehnert, Jens
Condurache, Alexandru Paul
author_facet Meiner, Lukas
Mehnert, Jens
Condurache, Alexandru Paul
contents To reduce the computational cost of convolutional neural networks (CNNs) on resource-constrained devices, structured pruning approaches have shown promise in lowering floating-point operations (FLOPs) without substantial drops in accuracy. However, most methods require fine-tuning or specific training procedures to achieve a reasonable trade-off between retained accuracy and reduction in FLOPs, adding computational overhead and requiring training data to be available. To this end, we propose HASTE (Hashing for Tractable Efficiency), a data-free, plug-and-play convolution module that instantly reduces a network's test-time inference cost without training or fine-tuning. Our approach utilizes locality-sensitive hashing (LSH) to detect redundancies in the channel dimension of latent feature maps, compressing similar channels to reduce input and filter depth simultaneously, resulting in cheaper convolutions. We demonstrate our approach on the popular vision benchmarks CIFAR-10 and ImageNet, where we achieve a 46.72% reduction in FLOPs with only a 1.25% loss in accuracy by swapping the convolution modules in a ResNet34 on CIFAR-10 for our HASTE module.
format Preprint
id arxiv_https___arxiv_org_abs_2309_17211
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Data-Free Dynamic Compression of CNNs for Tractable Efficiency
Meiner, Lukas
Mehnert, Jens
Condurache, Alexandru Paul
Computer Vision and Pattern Recognition
Machine Learning
To reduce the computational cost of convolutional neural networks (CNNs) on resource-constrained devices, structured pruning approaches have shown promise in lowering floating-point operations (FLOPs) without substantial drops in accuracy. However, most methods require fine-tuning or specific training procedures to achieve a reasonable trade-off between retained accuracy and reduction in FLOPs, adding computational overhead and requiring training data to be available. To this end, we propose HASTE (Hashing for Tractable Efficiency), a data-free, plug-and-play convolution module that instantly reduces a network's test-time inference cost without training or fine-tuning. Our approach utilizes locality-sensitive hashing (LSH) to detect redundancies in the channel dimension of latent feature maps, compressing similar channels to reduce input and filter depth simultaneously, resulting in cheaper convolutions. We demonstrate our approach on the popular vision benchmarks CIFAR-10 and ImageNet, where we achieve a 46.72% reduction in FLOPs with only a 1.25% loss in accuracy by swapping the convolution modules in a ResNet34 on CIFAR-10 for our HASTE module.
title Data-Free Dynamic Compression of CNNs for Tractable Efficiency
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2309.17211