Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kalle, Alper, Rudkiewicz, Theo, Ouerfelli, Mohamed-Oumar, Tamaazousti, Mohamed
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918197883568128
author Kalle, Alper
Rudkiewicz, Theo
Ouerfelli, Mohamed-Oumar
Tamaazousti, Mohamed
author_facet Kalle, Alper
Rudkiewicz, Theo
Ouerfelli, Mohamed-Oumar
Tamaazousti, Mohamed
contents Neural networks are widely used for image-related tasks but typically demand considerable computing power. Once a network has been trained, however, its memory- and compute-footprint can be reduced by compression. In this work, we focus on compression through tensorization and low-rank representations. Whereas classical approaches search for a low-rank approximation by minimizing an isotropic norm such as the Frobenius norm in weight-space, we use data-informed norms that measure the error in function space. Concretely, we minimize the change in the layer's output distribution, which can be expressed as $\lVert (W - \widetilde{W}) Σ^{1/2}\rVert_F$ where $Σ^{1/2}$ is the square root of the covariance matrix of the layer's input and $W$, $\widetilde{W}$ are the original and compressed weights. We propose new alternating least square algorithms for the two most common tensor decompositions (Tucker-2 and CPD) that directly optimize the new norm. Unlike conventional compression pipelines, which almost always require post-compression fine-tuning, our data-informed approach often achieves competitive accuracy without any fine-tuning. We further show that the same covariance-based norm can be transferred from one dataset to another with only a minor accuracy drop, enabling compression even when the original training dataset is unavailable. Experiments on several CNN architectures (ResNet-18/50, and GoogLeNet) and datasets (ImageNet, FGVC-Aircraft, Cifar10, and Cifar100) confirm the advantages of the proposed method.
format Preprint
id arxiv_https___arxiv_org_abs_2511_04494
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks
Kalle, Alper
Rudkiewicz, Theo
Ouerfelli, Mohamed-Oumar
Tamaazousti, Mohamed
Machine Learning
Computer Vision and Pattern Recognition
Neural networks are widely used for image-related tasks but typically demand considerable computing power. Once a network has been trained, however, its memory- and compute-footprint can be reduced by compression. In this work, we focus on compression through tensorization and low-rank representations. Whereas classical approaches search for a low-rank approximation by minimizing an isotropic norm such as the Frobenius norm in weight-space, we use data-informed norms that measure the error in function space. Concretely, we minimize the change in the layer's output distribution, which can be expressed as $\lVert (W - \widetilde{W}) Σ^{1/2}\rVert_F$ where $Σ^{1/2}$ is the square root of the covariance matrix of the layer's input and $W$, $\widetilde{W}$ are the original and compressed weights. We propose new alternating least square algorithms for the two most common tensor decompositions (Tucker-2 and CPD) that directly optimize the new norm. Unlike conventional compression pipelines, which almost always require post-compression fine-tuning, our data-informed approach often achieves competitive accuracy without any fine-tuning. We further show that the same covariance-based norm can be transferred from one dataset to another with only a minor accuracy drop, enabling compression even when the original training dataset is unavailable. Experiments on several CNN architectures (ResNet-18/50, and GoogLeNet) and datasets (ImageNet, FGVC-Aircraft, Cifar10, and Cifar100) confirm the advantages of the proposed method.
title Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.04494