A Hybrid Design for Low-Power Fault Tolerant One-Bit Full Adder for Neural Network Applications

C. Raji*, S. N. Prasad

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Artificial neural networks (ANNs) are, nowadays, used for implementing critical applications. However, ANN is not inherently fault tolerant. In the hardware implementation of constituent neuron comprising of adder and multiplier should be made fault tolerant, for the system to perform faultless. The building unit of a neuron or any arithmetic processing unit comprises of adders and by reduction in size of these devices, they are prone to transient errors. The power and area overhead in these implementations along with reliability have to be addressed. Here, a hybrid one-bit adder design is presented which resulted in reduction of 20% in number of transistors used and maintained the outputs at full swing thereby retaining the fidelity of the outputs. Power dissipated by the circuit is optimized to the extent which may be acceptable in any of the applications such as automotive, chip design, multimedia applications, and many more.

Original languageEnglish
Title of host publicationInternational Conference on Artificial Intelligence and Sustainable Engineering - Select Proceedings of AISE 2020
EditorsGoutam Sanyal, Carlos M. Travieso-González, Shashank Awasthi, Carla M. Pinto, B. R. Purushothama
PublisherSpringer Science and Business Media Deutschland GmbH
Pages277-293
Number of pages17
ISBN (Print)9789811685415
DOIs
Publication statusPublished - 2022
EventInternational Conference on Artificial Intelligence and Sustainable Engineering, AISE 2020 - Goa, India
Duration: 27-11-202029-11-2020

Publication series

NameLecture Notes in Electrical Engineering
Volume836
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

ConferenceInternational Conference on Artificial Intelligence and Sustainable Engineering, AISE 2020
Country/TerritoryIndia
CityGoa
Period27-11-2029-11-20

All Science Journal Classification (ASJC) codes

  • Industrial and Manufacturing Engineering

Fingerprint

Dive into the research topics of 'A Hybrid Design for Low-Power Fault Tolerant One-Bit Full Adder for Neural Network Applications'. Together they form a unique fingerprint.

Cite this