Output details

An application of a GA with Markov network surrogate to feature selection

This paper offers an original contribution to surrogate modelling. The Markov network approach offers an alternative the main surrogate methods which are radial basis functions, ANNs and Kriging. This is significant because it applies to discrete optimisation problems where surrogate modelling techniques are more difficult to apply. Extensive experimentation establishes the method as competitive and gives rigorous evidence on the various trade-offs involved in configuring the algorithm.