An application-specific array architecture for feedforward with backpropagation ANNs

Author

Malluhi, Q.M. ; Bayoumi, M.A. ; Rao, T.R.N.

Author_Institution

Center for Adv. Comput. Studies, Univ. of Southwestern Louisiana, Lafayette, LA, USA

fYear

1993

fDate

25-27 Oct 1993

Firstpage

333

Lastpage

344

Abstract

An application-specific array architecture for Artificial Neural Networks (ANNs) computation is proposed. This array is configured as a mesh-of-appendixed-trees (MAT). Algorithms to implement both the recall and the training phases of the multilayer feedforward with backpropagation ANN model are developed on MAT. The proposed MAT architecture requires only O(log N) time, while other reported techniques offer O(N) time, where N is the size of the largest layer. Beside the high speed performance, pipelining of more than one input pattern can be achieved which further improves the performance

Keywords

application specific integrated circuits; backpropagation; feedforward neural nets; neural chips; parallel architectures; pipeline processing; application-specific array architecture; backpropagation ANNs; feedforward; high speed performance; input pattern; mesh-of-appendixed-trees; multilayer feedforward; pipelining; training phases; Artificial neural networks; Backpropagation; Computational modeling; Computer architecture; Computer networks; Concurrent computing; Neural networks; Neurons; Parallel processing; Pipeline processing;

fLanguage

English

Publisher

ieee

Conference_Titel

Application-Specific Array Processors, 1993. Proceedings., International Conference on

Conference_Location

Venice

ISSN

1063-6862

Print_ISBN

0-8186-3492-8

Type

conf

DOI

10.1109/ASAP.1993.397156

Filename

397156