Computational Properties of Probabilistic Neural Networks

0350163 - ÚTIA 2011 RIV DE eng C - Conference Paper (international conference)
Grim, Jiří - Hora, Jan
Computational Properties of Probabilistic Neural Networks.
Artificial Neural Networks – ICANN 2010. Berlin Heidelberg: Springer Verlag, 2010 - (Diamantaras, K.; Duch, W.; Iliadis, L.), s. 31-40. Lecture Notes in Computer Science, LNCS, Volume 6354. ISBN 978-3-642-15818-6.
[ICANN 2010. International Conference on Artificial Neural Networks /20./. Thessaloniki (GR), 15.09.2010-18.09.2010]
R&D Projects: GA ČR GA102/07/1594; GA MŠMT 1M0572
Grant - others:GA MŠk(CZ) 2C06019
Institutional research plan: CEZ:AV0Z10750506
Keywords : Probabilistic neural networks * Statistical pattern recognition * Subspace approach * Overfitting reduction
Subject RIV: IN - Informatics, Computer Science
http://library.utia.cas.cz/separaty/2010/RO/grim-computational properties of probabilistic neural networks.pdf

We discuss the problem of overfitting of probabilistic neural networks in the framework of statistical pattern recognition. The probabilistic approach to neural networks provides a statistically justified subspace method of classification. The underlying structural mixture model includes binary structural parameters and can be optimized by EM algorithm in full generality. Formally, the structural model reduces the number of parameters included and therefore the structural mixtures become less complex and less prone to overfitting. We illustrate how recognition accuracy and the effect of overfitting is influenced by mixture complexity and by the size of training data set.
Permanent Link: http://hdl.handle.net/11104/0190237