The effects of limited precision implementation on the performance of radial basis functions

Author

Sen, Selcuk ; Robertson, William ; Phillips, William

Author_Institution

Dept. of Electr. Eng., Tech. Univ. Nova Scotia, Halifax, NS, Canada

Volume

1

fYear

1997

fDate

25-28 May 1997

Firstpage

149

Abstract

The authors present the effects of employing limited precision arithmetic on radial basis function networks (RBF). A statistical formula is first derived to obtain an analytical expression for the error so that it can be used for hand-calculations. A speech processing example is chosen to compare the theoretical results with its simulation counterparts. The results indicate that limited precision implementation achieves as nearly good recognition results as its floating counterpart

Keywords

digital arithmetic; error analysis; feedforward neural nets; speech recognition; statistical analysis; analytical expression; error; limited precision arithmetic; performance; radial basis function networks; simulation; speech processing; speech recognition; statistical formula; Clustering algorithms; Equations; Euclidean distance; Floating-point arithmetic; Joining processes; Least squares approximation; Measurement standards; Radial basis function networks; Speech processing; Supervised learning;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrical and Computer Engineering, 1997. Engineering Innovation: Voyage of Discovery. IEEE 1997 Canadian Conference on

Conference_Location

St. Johns, Nfld.

ISSN

0840-7789

Print_ISBN

0-7803-3716-6

Type

conf

DOI

10.1109/CCECE.1997.614812

Filename

614812