Computational constraints often limit the practical applicability of coherent Bayes solutions to unsupervised sequential learning problems. These problems arise when attemps are made to learn about parameters on the basis of unclassified observations., each stemming from any one of k cases (k ≥ 2).
In this paper, the difficulties of the Bayes process will be discussed and existing approximate learning procedures will be reviewed for broad types of problems involving mixtures of probability density. In particular a quasi-Bayes approximate learning procedure will be motivated and defined and its convergence properties will be reported for several special cases.
@article{urn:eudml:doc:40812,
title = {Approximation of unsupervised Bayes learning procedures.},
journal = {Trabajos de Estad\'\i stica e Investigaci\'on Operativa},
volume = {31},
year = {1980},
pages = {69-81},
language = {en},
url = {http://dml.mathdoc.fr/item/urn:eudml:doc:40812}
}
Makov, Udi E. Approximation of unsupervised Bayes learning procedures.. Trabajos de Estadística e Investigación Operativa, Tome 31 (1980) pp. 69-81. http://gdmltest.u-ga.fr/item/urn:eudml:doc:40812/