Application of the piecewise linear approximation method in a hardware accelerators of a neural networks based on a reconfigurable computing environments
This paper considers the application of piecewise linear approximation in hardware accelerators of neural networks built according to the concept of reconfigurable computing environments (RCE). A method for providing an uneven approximation step is shown. The results of the application of the propos...
| Опубликовано в: : | Distributed computer and communication networks : 25th International conference, DCCN 2022, Moscow, Russia, September 26-29, 2022 : revised selected papers P. 63-74 |
|---|---|
| Главный автор: | |
| Другие авторы: | |
| Формат: | Статья в сборнике |
| Язык: | English |
| Предметы: | |
| Online-ссылка: | https://vital.lib.tsu.ru/vital/access/manager/Repository/koha:001156737 Перейти в каталог НБ ТГУ |
| Итог: | This paper considers the application of piecewise linear approximation in hardware accelerators of neural networks built according to the concept of reconfigurable computing environments (RCE). A method for providing an uneven approximation step is shown. The results of the application of the proposed approaches for the approximation of exponential and sigmoid functions actively used in machine learning algorithms are presented. The use of RCE allows the development of hardware algorithms for information processing, characterized by high performance characteristics due to the architectural properties of RCE. |
|---|---|
| Библиография: | Библиогр.: 14 назв. |
| ISBN: | 9783031306471 |