Dergiler / Turkish Journal of Engineering and Environmental Sciences / 1998 / Cilt: 22 - Sayı: 2

A Study on a Learning Algorithm

Sayfa
109–116
DOI
—

Özet

In this paper, an algorithm which can generate a nonlineer multi-input function through a learning process is studied. This algorithm, used by the Sulzer A.G. company for building heating automation, has apparent potentials for use in control and diagnostics. For this reason, some of its fundamental properties like accuracy, stability and convergence rate which were considered to be important regarding practical design were investigated, and the results are reported here.

Abstract

In this paper, an algorithm which can generate a nonlineer multi-input function through a learning process is studied. This algorithm, used by the Sulzer A.G. company for building heating automation, has apparent potentials for use in control and diagnostics. For this reason, some of its fundamental properties like accuracy, stability and convergence rate which were considered to be important regarding practical design were investigated, and the results are reported here.

Anahtar kelimeler: Learning systems, Trainable Function Generator, Nonlinear Function Generator.