Evolving the Topology of Hidden Markov Models Using Evolutionary Algorithms

Hidden Markov models (HMM) are widely used for speech recognition and have recently gained a lot of attention in the bioinformatics community, because of their ability to capture the information buried in biological sequences. Usually, heuristic algorithms such as Baum-Welch are used to estimate the model parameters. However, Baum-Welch has a tendency to stagnate on local optima. Furthermore, designing an optimal HMM topology usually requires a priori knowledge from a field expert and is usually found by trial-and-error. In this study, we present an evolutionary algorithm capable of evolving both the topology and the model parameters of HMMs. The applicability of the method is exemplified on a secondary structure prediction problem.