A new approach to the analysis of Mira light curvesTwo different but complementary methods for predicting Mira luminosities are presented. One method is derived from a Fourier analysis, it requires performing deconvolution, and its results are not certain due to the inherent instability of deconvolution problems. The other method is a learning method utilizing artificial intelligence techniques where a light curve is presented as an ordered sequence of pseudocycles, and rules are learned by linking the characteristics of several consecutive pseudocycles to one characteristic of the future cycle. It is observed that agreement between these methods is obtainable when it is possible to eliminate similar false frequencies from the preliminary power spectrum and to improve the degree of confidence in the rules.
Document ID
19910044299
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Mennessier, M. O. (Centre National de la Recherche Scientifique Montpellier, France)
Barthes, D. (CNRS; Montpellier II, Universite France)
Mattei, J. A. (American Association of Variable Star Observers Cambridge, MA, United States)
Date Acquired
August 14, 2013
Publication Date
January 1, 1990
Subject Category
Astrophysics
Meeting Information
Meeting: Confrontation between Stellar Pulsation and Evolution