An investigation of high-order bidirectional associative memories: performance, applications, and parallel implementations

Author

Wu, Chwan-Hwa ; Wang, Chia-Jiu ; Tai, Heng-Ming ; Roland, David A.

Author_Institution

Dept. of Electr., Auburn Univ., AL, USA

fYear

1990

fDate

1-3 May 1990

Firstpage

495

Abstract

An investigation of a bidirectional associative memory (BAM) incorporating the high-order nonlinearity is presented. The improvement of the memory storage capacity and the error-correcting capability of the high-order bidirectional associative memory (HOBAM) is discussed. The application of a frequency classifier is presented, and the performance of HOBAM and BAM are compared. The HOBAM is superior to the BAM in classifying cluttered signals. The application of image classification is reported. The parallel processing of HOBAM on an Allian/FX80 for frequency classification and on a dynamically reconfigurable multiprocessor system for image recognition is discussed

Keywords

computerised picture processing; computerised signal processing; content-addressable storage; error correction; neural nets; parallel algorithms; Allian/FX80; CAM; bidirectional associative memories; dynamically reconfigurable multiprocessor system; error-correcting capability; frequency classifier; high-order nonlinearity; image classification; image recognition; memory storage capacity; parallel implementations; parallel processing; Application software; Associative memory; Frequency; Image recognition; Magnesium compounds; Multiprocessing systems; Neurons; Parallel processing; Springs; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1990., IEEE International Symposium on

Conference_Location

New Orleans, LA

Type

conf

DOI

10.1109/ISCAS.1990.112095

Filename

112095