Approximate arithmetic for modern neural networks and FPGAs

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21240%2F22%3A00358459" target="_blank" >RIV/68407700:21240/22:00358459 - isvavai.cz</a>

Result on the web

<a href="https://doi.org/10.1109/MECO55406.2022.9797141" target="_blank" >https://doi.org/10.1109/MECO55406.2022.9797141</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/MECO55406.2022.9797141" target="_blank" >10.1109/MECO55406.2022.9797141</a>

Result language

angličtina

Original language name

Approximate arithmetic for modern neural networks and FPGAs

Original language description

Approximate arithmetic is a very important approach for implementing neural networks in embedded hardware. The requirements of real-time applications with respect to the size of deep neural networks force designers to simplify the neural processing elements. Not only the reduction of precision of model parameters to a few bits, but also the use of approximate arithmetic increases computational power and saves on-chip resources beyond exact computation. Since it was shown that linearly approximated functions are suitable for implementing neural networks in hardware, FPGAs have improved. So we decided to re-implement the processing element on a modern FPGA and to present implementation results regarding speed and resource consumption. A neural processing element based on linearly approximated functions was implemented in Vivado and tested on an xc7 FPGA. The results show that the architecture saves significant resources and a clock frequency above 100 MHz can be achieved in pipelined design.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

—

Continuities

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year

2022

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Proceedings of the 11th Mediterranean Conference on Embedded Computing (MECO 2022)

ISBN

978-1-6654-6828-2

ISSN

2377-5475

e-ISSN

—

Number of pages

4

Pages from-to

351-354

Publisher name

Institute of Electrical and Electronics Engineers, Inc.

Place of publication

—

Event location

Budva

Event date

Jun 7, 2022

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

000855969800072