Kybernetika 39 no. 5, 511-520, 2003

Neural networks using Bayesian training

Gabriela Andrejková and Miroslav Levický


Bayesian probability theory provides a framework for data modeling. In this framework it is possible to find models that are well-matched to the data, and to use these models to make nearly optimal predictions. In connection to neural networks and especially to neural network learning, the theory is interpreted as an inference of the most probable parameters for the model and the given training data. This article describes an application of Neural Networks using the Bayesian training to the problem of Predictions of Geomagnetic Storms.