Skip to main navigation Skip to search Skip to main content

Piecewise-linear modeling of analog circuits based on model extraction from trained neural networks

  • Stony Brook University
  • Hofstra University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

6 Scopus citations

Abstract

This paper presents a new technique for automatically creating analog circuit models. The method extracts piecewise linear models from trained neural networks. A model is a set of linear dependencies between circuit performance and design parameters. The paper illustrates the technique for an OTA circuit for which models for gain and bandwidth are generated. As experiments show, the obtained models have simple form that accurately fits the sampled points. These models are useful for fast simulation of systems with nonlinear behavior and performance.

Original languageEnglish
Title of host publicationBMAS 2002 - Proceedings of the 2002 IEEE International Workshop on Behavioral Modeling and Simulation
PublisherIEEE Computer Society
Pages41-46
Number of pages6
ISBN (Electronic)078037634X
DOIs
StatePublished - 2002
EventIEEE International Workshop on Behavioral Modeling and Simulation, BMAS 2002 - Santa Rosa, United States
Duration: Oct 6 2002Oct 8 2002

Publication series

NameProceedings of the IEEE International Workshop on Behavioral Modeling and Simulation, BMAS
Volume2002-January
ISSN (Print)2160-3804
ISSN (Electronic)2160-3812

Conference

ConferenceIEEE International Workshop on Behavioral Modeling and Simulation, BMAS 2002
Country/TerritoryUnited States
CitySanta Rosa
Period10/6/0210/8/02

Keywords

  • Analog circuits
  • Analog computers
  • Circuit synthesis
  • Computer networks
  • Design engineering
  • Laboratories
  • Neural networks
  • Piecewise linear techniques
  • Predictive models
  • Very large scale integration

Fingerprint

Dive into the research topics of 'Piecewise-linear modeling of analog circuits based on model extraction from trained neural networks'. Together they form a unique fingerprint.

Cite this