Neural system design with the integrated neurocomputing architecture

Author

Mukai, Paul ; Busa, Mark ; Kazlas, Peter

Author_Institution

Charles Stark Draper Lab. Inc., Cambridge, MA, USA

fYear

1993

fDate

5-6 Mar 1993

Firstpage

32

Lastpage

36

Abstract

Design and implementation issues of high-performance VLSI systems in the form of deployable neural networks for pattern recognition applications are addressed, in particular the integrated neurocomputing architecture (INCA), which was developed with a complete system approach involving the integration of custom analog IC design, digital and analog board-level design, neural network development software, and application-specific hardware and software elements, is considered. The design methodology is discussed to demonstrate the capability of INCA as a usable neuroprocessing system. It serves as an example of a system that highlights many design automation issues for future mixed-signal, highly integrated systems

Keywords

VLSI; neural chips; parallel architectures; pattern recognition equipment; INCA; VLSI systems; board-level design; custom analog IC design; deployable neural networks; design automation issues; integrated neurocomputing architecture; neural network development software; neuroprocessing system; pattern recognition applications; Analog integrated circuits; Application software; Application specific integrated circuits; Computer architecture; Design methodology; Digital integrated circuits; Neural network hardware; Neural networks; Pattern recognition; Very large scale integration;

fLanguage

English

Publisher

ieee

Conference_Titel

VLSI, 1993. 'Design Automation of High Performance VLSI Systems', Proceedings., Third Great Lakes Symposium on

Conference_Location

Kalamazoo, MI

Print_ISBN

0-8186-3430-8

Type

conf

DOI

10.1109/GLSV.1993.224486

Filename

224486