You are here

Optimization of a Parallel Cordic Architecture to Compute the Gaussian Potential Function in Neural Networks

Title: Optimization of a Parallel Cordic Architecture to Compute the Gaussian Potential Function in Neural Networks.
30 views
7 downloads
Name(s): Chandrasekhar, Nanditha, author
Baese, Anke Meyer, professor directing thesis
Baese, Uwe Meyer, committee member
Foo, Simon, committee member
Department of Electrical and Computer Engineering, degree granting department
Florida State University, degree granting institution
Type of Resource: text
Genre: Text
Issuance: monographic
Date Issued: 2005
Publisher: Florida State University
Place of Publication: Tallahassee, Florida
Physical Form: computer
online resource
Extent: 1 online resource
Language(s): English
Abstract/Description: Many pattern recognition tasks employ artificial neural networks based on radial basis functions. The statistical characteristics of pattern generating processes are determined by neural networks. The Gaussian potential function is the most common radial basis function considered which includes square and exponential function calculations. The Coordinate Rotations Digital Computer, CORDIC algorithm which is used to compute the exponential function and the exponent was first derived by Volder in 1959 for calculating trigonometric functions and conversions between rectangular and polar co-ordinates. It was later developed by Walther, the CORDIC is a class of shift-add algorithms for rotating vectors in a plane. In a nutshell, the CORDIC rotator performs a rotation using a series of specific incremental rotation angles selected so that each is performed by a shift and add operation. This thesis focuses on implementation of new parallel hardware architecture to compute the Gaussian Potential Function in neural basis classifiers for pattern recognition. The new hardware proposed computes the exponential function and the exponent simultaneously in parallel thus reducing computational delay in the output function. The new CORDIC is synthesized by Altera's MAX PLUS II software for FLEX 10 K device and improvised for calculation of Radix 4. Case studies are presented and compared on the performance of Radix 2 and Radix 4 design based on the speed and the size occupied respectively. It is observed that though the area occupied by Radix 4 is more as compared to Radix 2 there is speed improvement which is desirable.
Identifier: FSU_migr_etd-3905 (IID)
Submitted Note: A Thesis submitted to the Department of Electrical and Computer Engineering in partial fulfillment of the requirements for the degree of Master of Science.
Degree Awarded: Spring Semester, 2005.
Date of Defense: April 7, 2005.
Keywords: Gaussian potential function, Cordic
Bibliography Note: Includes bibliographical references.
Advisory Committee: Anke Meyer Baese, Professor Directing Thesis; Uwe Meyer Baese, Committee Member; Simon Foo, Committee Member.
Subject(s): Electrical engineering
Computer engineering
Persistent Link to This Record: http://purl.flvc.org/fsu/fd/FSU_migr_etd-3905
Owner Institution: FSU

Choose the citation style.
Chandrasekhar, N. (2005). Optimization of a Parallel Cordic Architecture to Compute the Gaussian Potential Function in Neural Networks. Retrieved from http://purl.flvc.org/fsu/fd/FSU_migr_etd-3905