Modular search space for automated design of neural architecture

Скорочена інформація про документ
dc.contributor.authorRadiuk, P.M.
dc.date.accessioned2021-12-14T11:39:16Z
dc.date.available2021-12-14T11:39:16Z
dc.date.issued2020-12-14
dc.descriptionhttps://doi.org/10.33243/2518-7139-2020-1-1-37-44uk_UA
dc.description.abstractThe past years of research have shown that automated machine learning and neural architecture search are an inevitable future for image recognition tasks. In addition, a crucial aspect of any automated search is the predefined search space. As many studies have demonstrated, the modularization technique may simplify the underlying search space by fostering successful blocks’ reuse. In this regard, the presented research aims to investigate the use of modularization in automated machine learning. In this paper, we propose and examine a modularized space based on the substantial limitation to seeded building blocks for neural architecture search. To make a search space viable, we presented all modules of the space as multisectoral networks. Therefore, each architecture within the search space could be unequivocally described by a vector. In our case, a module was a predetermined number of parameterized layers with information about their relationships. We applied the proposed modular search space to a genetic algorithm and evaluated it on the CIFAR-10 and CIFAR-100 datasets based on modules from the NAS-Bench-201 benchmark. To address the complexity of the search space, we randomly sampled twenty-five modules and included them in the database. Overall, our approach retrieved competitive architectures in averaged 8 GPU hours. The final model achieved the validation accuracy of 89.1% and 73.2% on the CIFAR-10 and CIFAR- 100 datasets, respectively. The learning process required slightly fewer GPU hours compared to other approaches, and the resulting network contained fewer parameters to signal lightness of the model. Such an outcome may indicate the considerable potential of sophisticated ranking approaches. The conducted experiments also revealed that a straightforward and transparent search space could address the challenging task of neural architecture search. Further research should be undertaken to explore how the predefined knowledge base of modules could benefit modular search space.uk_UA
dc.identifier.citationRadiuk P. M. Modular search space for automated design of neural architecture // Proceedings of the O.S. Popov ОNAT. 2020. Vol. 1, No 1. P. 37-44. https://doi.org/10.33243/2518-7139-2020-1-1-37-44.uk_UA
dc.identifier.urihttps://elar.khmnu.edu.ua/handle/123456789/11051
dc.language.isoenuk_UA
dc.publisherProceedings of the O.S. Popov ОNATuk_UA
dc.subjectsearch spaceuk_UA
dc.subjectmodularizationuk_UA
dc.subjectautomluk_UA
dc.subjectneural architecture searchuk_UA
dc.subjectgenetic algorithmuk_UA
dc.subject.udc004.023+519.876.5uk_UA
dc.titleModular search space for automated design of neural architectureuk_UA
dc.typeСтаттяuk_UA
Файли
Контейнер файлів
Зараз показуємо 1 - 1 з 1
Ескіз
Назва:
1165-Текст статьи-2416-1-10-20210129.pdf
Розмір:
316.11 KB
Формат:
Adobe Portable Document Format
Опис:
Завантажити
Ліцензійна угода
Зараз показуємо 1 - 1 з 1
Назва:
license.txt
Розмір:
4.26 KB
Формат:
Item-specific license agreed upon to submission
Опис:
Завантажити
Зібрання
Кафедра комп’ютерних наук