Bayesian learning on dependent features
Eric Bengfort · 2011
Bayesian Learning is a robust machine learning algorithm. In addition to working well with limited sample sizes, Bayesian Learning also provides a probabilistic categorization. The caveat of performing a Bayesian Analysis is that the features of Training Instances are typically assumed to be independent of each other. As a result, Bayesian Learning will struggle to properly classify Training Instances with dependent features. This paper proposes an extension to the Bayesian Learning Algorithm which is capable of overcoming this weakness. This newly proposed algorithm will create new candidates for Bayesian Classification by performing permutations upon the Original Training Data. It is hoped that this upstart algorithm will be applied to biological data sets to find the solution to an unsolved medical problem.