Selecting feature-based models
Amir Jalalirad, T.J. Tjalkens, Jean‐Paul M. G. Linnartz, S. Pollin, L. Van der Perre, A. Stas · TU/e Research Portal · 2013
In a classification problem, we would like to assign a model to the observed data using its features. If the number of these features is large, considering dependencies between them to come up with an appropriate model becomes a challenge. In this paper an efficient method is introduced for selecting a model that fits the observed data sequence with a large number of features. This method is constructed by a modification of an earlier studied algorithm. It is shown that the model selected by this method has the Maximum Likelihood probability and can capture the dependencies between the features of the data sequence. We prove that through this method, the algebraic workload required for calculating the Maximum Likelihood model is reduced.