Title

Multi-stream Extended Kalman Filter Training of Neural Networks on a Simd Parallel Machine

Abstract

The extended Kalman filter (EKF) algorithm has been shown to be advantageous for neural network trainings. This paper presents a method to do the EFK training on a SIMB parallel machine. We use multi-stream decoupled extended Kalman filter (DEKF) training algorithm which can provide more improved trained network weights and efficient use of the parallel resource. The performance of the parallel DEKF training algorithm is studied and simulation results for the estimation of the wind power using neural networks are provided.

Department(s)

Electrical and Computer Engineering

Sponsor(s)

Central and South West Services
Electric Power Research Institution
National Science Foundation (U.S.)
Zond, Inc.

Keywords and Phrases

Artificial Intelligence; Neural Networks

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 1999 American Society of Mechanical Engineers (ASME), All rights reserved.

This document is currently not available here.


Share

 
COinS