Activated Current Sensing Circuit for Resistive Neuromorphic Networks

Mohammed E. Fouda, Ahmed M. Eltawil, Fadi Kurdahi

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

1 Scopus citations

Abstract

Recently, resistive-based neural networks have been adopted to build deep learning architectures, where the small area footprint RRAMs (memristors) enables unprecedentedly large neural networks. In this work, we introduce a current sensing circuit and an integrated activation function for resistive neural networks for the first time. This circuit is vital since it is replicated hundreds of times at the outputs of the neurons and thus should be low power and ultra-compact. The proposed circuit is designed using TSMC65nm. The circuit is based on the current conveyor principle and a simple inverter to create the required activation function. The obtained response is curve-fitted to hyperbolic tangent and sigmoid functions to get the accurate expression for the nonlinear function used to design the training technique of the entire neural network. Finally, the proposed circuit is tested in a four-bit neural network based ADC.
Original languageEnglish (US)
Title of host publication17th IEEE International New Circuits and Systems Conference, NEWCAS 2019
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Print)9781728110318
DOIs
StatePublished - Jun 1 2019
Externally publishedYes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2020-03-18

Fingerprint Dive into the research topics of 'Activated Current Sensing Circuit for Resistive Neuromorphic Networks'. Together they form a unique fingerprint.

Cite this