Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks

Neural networks have become a cornerstone of computer vision applications, with tasks ranging from image classification to object detection. However, challenges such as hyperparameter optimization (HPO) and model compression remain critical for improving performance and deploying models on resource-...

Full description

Saved in:
Bibliographic Details
Main Authors: A. Naumov, A. Melnikov, M. Perelshtein, Ar. Melnikov, V. Abronin, F. Oksanichenko
Format: Article
Language:English
Published: MDPI AG 2025-02-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/4/1852
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850081161546563584
author A. Naumov
A. Melnikov
M. Perelshtein
Ar. Melnikov
V. Abronin
F. Oksanichenko
author_facet A. Naumov
A. Melnikov
M. Perelshtein
Ar. Melnikov
V. Abronin
F. Oksanichenko
author_sort A. Naumov
collection DOAJ
description Neural networks have become a cornerstone of computer vision applications, with tasks ranging from image classification to object detection. However, challenges such as hyperparameter optimization (HPO) and model compression remain critical for improving performance and deploying models on resource-constrained devices. In this work, we address these challenges using Tensor Network-based methods. For HPO, we propose and evaluate the TetraOpt algorithm against various optimization algorithms. These evaluations were conducted on subsets of the NATS-Bench dataset, including CIFAR-10, CIFAR-100, and ImageNet subsets. TetraOpt consistently demonstrated superior performance, effectively exploring the global optimization space and identifying configurations with higher accuracies. For model compression, we introduce a novel iterative method that combines CP, SVD, and Tucker tensor decompositions. Applied to ResNet-18 and ResNet-152, we evaluated our method on the CIFAR-10 and Tiny ImageNet datasets. Our method achieved compression ratios of up to <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>14.5</mn><mo>×</mo></mrow></semantics></math></inline-formula> for ResNet-18 and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>2.5</mn><mo>×</mo></mrow></semantics></math></inline-formula> for ResNet-152. Additionally, the inference time for processing an image on a CPU remained largely unaffected, demonstrating the practicality of the method.
format Article
id doaj-art-e4634d42db29463ab54a17c63d8a58cf
institution DOAJ
issn 2076-3417
language English
publishDate 2025-02-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-e4634d42db29463ab54a17c63d8a58cf2025-08-20T02:44:47ZengMDPI AGApplied Sciences2076-34172025-02-01154185210.3390/app15041852Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural NetworksA. Naumov0A. Melnikov1M. Perelshtein2Ar. Melnikov3V. Abronin4F. Oksanichenko5Terra Quantum AG, Kornhausstrasse 25, 9000 St. Gallen, SwitzerlandTerra Quantum AG, Kornhausstrasse 25, 9000 St. Gallen, SwitzerlandTerra Quantum AG, Kornhausstrasse 25, 9000 St. Gallen, SwitzerlandTerra Quantum AG, Kornhausstrasse 25, 9000 St. Gallen, SwitzerlandTerra Quantum AG, Kornhausstrasse 25, 9000 St. Gallen, SwitzerlandTerra Quantum AG, Kornhausstrasse 25, 9000 St. Gallen, SwitzerlandNeural networks have become a cornerstone of computer vision applications, with tasks ranging from image classification to object detection. However, challenges such as hyperparameter optimization (HPO) and model compression remain critical for improving performance and deploying models on resource-constrained devices. In this work, we address these challenges using Tensor Network-based methods. For HPO, we propose and evaluate the TetraOpt algorithm against various optimization algorithms. These evaluations were conducted on subsets of the NATS-Bench dataset, including CIFAR-10, CIFAR-100, and ImageNet subsets. TetraOpt consistently demonstrated superior performance, effectively exploring the global optimization space and identifying configurations with higher accuracies. For model compression, we introduce a novel iterative method that combines CP, SVD, and Tucker tensor decompositions. Applied to ResNet-18 and ResNet-152, we evaluated our method on the CIFAR-10 and Tiny ImageNet datasets. Our method achieved compression ratios of up to <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>14.5</mn><mo>×</mo></mrow></semantics></math></inline-formula> for ResNet-18 and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>2.5</mn><mo>×</mo></mrow></semantics></math></inline-formula> for ResNet-152. Additionally, the inference time for processing an image on a CPU remained largely unaffected, demonstrating the practicality of the method.https://www.mdpi.com/2076-3417/15/4/1852tensortrain optimisationhyperparameter optimisationimage classificationcomputer vision and pattern recognitionconvolutional neural networksmodel compression
spellingShingle A. Naumov
A. Melnikov
M. Perelshtein
Ar. Melnikov
V. Abronin
F. Oksanichenko
Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
Applied Sciences
tensortrain optimisation
hyperparameter optimisation
image classification
computer vision and pattern recognition
convolutional neural networks
model compression
title Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
title_full Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
title_fullStr Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
title_full_unstemmed Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
title_short Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
title_sort tensor network methods for hyperparameter optimization and compression of convolutional neural networks
topic tensortrain optimisation
hyperparameter optimisation
image classification
computer vision and pattern recognition
convolutional neural networks
model compression
url https://www.mdpi.com/2076-3417/15/4/1852
work_keys_str_mv AT anaumov tensornetworkmethodsforhyperparameteroptimizationandcompressionofconvolutionalneuralnetworks
AT amelnikov tensornetworkmethodsforhyperparameteroptimizationandcompressionofconvolutionalneuralnetworks
AT mperelshtein tensornetworkmethodsforhyperparameteroptimizationandcompressionofconvolutionalneuralnetworks
AT armelnikov tensornetworkmethodsforhyperparameteroptimizationandcompressionofconvolutionalneuralnetworks
AT vabronin tensornetworkmethodsforhyperparameteroptimizationandcompressionofconvolutionalneuralnetworks
AT foksanichenko tensornetworkmethodsforhyperparameteroptimizationandcompressionofconvolutionalneuralnetworks