Transverse-layer partitioning of artificial neural networks for image classification
We discuss issues of modular learning in artificial neural networks and explore possibilities of the partial use of modules when the computational resources are limited. The proposed method is based on the ability of a wavelet transform to separate information into high- and low-frequency parts. Usi...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Samara National Research University
2024-04-01
|
| Series: | Компьютерная оптика |
| Subjects: | |
| Online Access: | https://www.computeroptics.ru/eng/KO/Annot/KO48-2/480218e.html |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832540914434703360 |
|---|---|
| author | N.A. Vershkov M.G. Babenko N.N. Kuchukov V.A. Kuchukov N.N. Kucherov |
| author_facet | N.A. Vershkov M.G. Babenko N.N. Kuchukov V.A. Kuchukov N.N. Kucherov |
| author_sort | N.A. Vershkov |
| collection | DOAJ |
| description | We discuss issues of modular learning in artificial neural networks and explore possibilities of the partial use of modules when the computational resources are limited. The proposed method is based on the ability of a wavelet transform to separate information into high- and low-frequency parts. Using the expertise gained in developing convolutional wavelet neural networks, the authors perform a transverse-layer partitioning of the network into modules for the further partial use on devices with low computational capability. The theoretical justification of this approach in the paper is supported by experimentally dividing the MNIST database into 2 and 4 modules before using them sequentially and measuring the respective accuracy and performance. When using the individual modules, a two-fold (or higher) performance gain is achieved. The theoretical statements are verified using an AlexNet-like network on the GTSRB dataset, with a performance gain of 33% per module with no loss of accuracy. |
| format | Article |
| id | doaj-art-bc198cc6abc04e5b87da6aba8980eadd |
| institution | Kabale University |
| issn | 0134-2452 2412-6179 |
| language | English |
| publishDate | 2024-04-01 |
| publisher | Samara National Research University |
| record_format | Article |
| series | Компьютерная оптика |
| spelling | doaj-art-bc198cc6abc04e5b87da6aba8980eadd2025-02-04T13:02:04ZengSamara National Research UniversityКомпьютерная оптика0134-24522412-61792024-04-0148231232010.18287/2412-6179-CO-1278Transverse-layer partitioning of artificial neural networks for image classificationN.A. Vershkov0M.G. Babenko1N.N. Kuchukov2V.A. Kuchukov3N.N. Kucherov4North-Caucasus Center for Mathematical Research, North Caucasus Federal UniversityNorth-Caucasus Center for Mathematical Research, North Caucasus Federal UniversityNorth-Caucasus Center for Mathematical Research, North Caucasus Federal UniversityNorth-Caucasus Center for Mathematical Research, North Caucasus Federal UniversityNorth-Caucasus Center for Mathematical Research, North Caucasus Federal UniversityWe discuss issues of modular learning in artificial neural networks and explore possibilities of the partial use of modules when the computational resources are limited. The proposed method is based on the ability of a wavelet transform to separate information into high- and low-frequency parts. Using the expertise gained in developing convolutional wavelet neural networks, the authors perform a transverse-layer partitioning of the network into modules for the further partial use on devices with low computational capability. The theoretical justification of this approach in the paper is supported by experimentally dividing the MNIST database into 2 and 4 modules before using them sequentially and measuring the respective accuracy and performance. When using the individual modules, a two-fold (or higher) performance gain is achieved. The theoretical statements are verified using an AlexNet-like network on the GTSRB dataset, with a performance gain of 33% per module with no loss of accuracy.https://www.computeroptics.ru/eng/KO/Annot/KO48-2/480218e.htmlwavelet transformartificial neural networksconvolutional layerorthogonal transformsmodular learning |
| spellingShingle | N.A. Vershkov M.G. Babenko N.N. Kuchukov V.A. Kuchukov N.N. Kucherov Transverse-layer partitioning of artificial neural networks for image classification Компьютерная оптика wavelet transform artificial neural networks convolutional layer orthogonal transforms modular learning |
| title | Transverse-layer partitioning of artificial neural networks for image classification |
| title_full | Transverse-layer partitioning of artificial neural networks for image classification |
| title_fullStr | Transverse-layer partitioning of artificial neural networks for image classification |
| title_full_unstemmed | Transverse-layer partitioning of artificial neural networks for image classification |
| title_short | Transverse-layer partitioning of artificial neural networks for image classification |
| title_sort | transverse layer partitioning of artificial neural networks for image classification |
| topic | wavelet transform artificial neural networks convolutional layer orthogonal transforms modular learning |
| url | https://www.computeroptics.ru/eng/KO/Annot/KO48-2/480218e.html |
| work_keys_str_mv | AT navershkov transverselayerpartitioningofartificialneuralnetworksforimageclassification AT mgbabenko transverselayerpartitioningofartificialneuralnetworksforimageclassification AT nnkuchukov transverselayerpartitioningofartificialneuralnetworksforimageclassification AT vakuchukov transverselayerpartitioningofartificialneuralnetworksforimageclassification AT nnkucherov transverselayerpartitioningofartificialneuralnetworksforimageclassification |