A fast and low-power distance computation unit dedicated to neural networks, based on redundant arithmetic

Abstract : This paper presents the design of a fast and low power consumption distance computation unit : /spl Sigma//sub i/(A/sub i/-B/sub i/)/sup 2/. It is dedicated to the digital RBF neural network implementation. The proposed architecture is composed of two parts. The first computes the distance (A/sub i/-B/sub i/)/sup 2/, and the second performs the sum of these distances. It is based on an efficient squarer in redundant arithmetic. Thank to this operator, the distance measure circuits developed offer better performances than those based on classical arithmetic. The average gain is equal to 11% in delay and 18% in power consumption.
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01571016
Contributor : Lip6 Publications <>
Submitted on : Tuesday, August 1, 2017 - 2:17:27 PM
Last modification on : Thursday, March 21, 2019 - 1:13:30 PM

Identifiers

Citation

Yannick Dumonteix, Yann Bajot, Habib Mehrez. A fast and low-power distance computation unit dedicated to neural networks, based on redundant arithmetic. IEEE International Symposium on Circuits and Systems (ISCAS'2001), May 2001, Sydney, Australia. pp.878-881, ⟨10.1109/ISCAS.2001.922378⟩. ⟨hal-01571016⟩

Share

Metrics

Record views

58