|
ИСТИНА |
Войти в систему Регистрация |
ФНКЦ РР |
||
Noise data impact on catastrophic interference in spiking neural networksстатья
- Автор: Antonov Dmitry I.
- Сборник: THE VIII INTERNATIONAL YOUNG RESEARCHERS’ CONFERENCE – PHYSICS, TECHNOLOGY, INNOVATIONS (PTI-2021)
- Год издания: 2022
- Место издания: AIP Publishing
- DOI: 10.1063/5.0088710
- Аннотация: The problem of catastrophic interference in Spiking Neural Networks (SNNs) is investigated. We considered the situation of continuous learning without access to an initial dataset used to train the model. The pseudo-rehearsal method is investigated as one of the ways to overcome catastrophic interference. The impact of noise images on the training process and SNNs’ recognition ability is studied.
- Добавил в систему: Антонов Дмитрий Иванович