Comparison of discrete and continuous bio-inspired metaheuristics for HMM likelihood maximization
Sebastien Aupetit, Nicolas Monmarch, Mohamed SlimaneAbstract: In this paper, we present a comparative analysis of three population-based and bio-inspired metaheuristics for the Hidden Markov Model (HMM) likelihood maximization problem. After a short description of HMMs and the corresponding continuous search space, we introduce three population based metaheuristics used to solve the optimization problem: a genetic algorithm, an ant based algorithm and a particle swarm optimization method. The third part of the paper is devoted to the analysis of similarities and differences between algorithms both theoretically and experimentally.
Keywords:
Hidden Markov Model, Likelihood maximization, Genetic algorithm, Ant algorithm, The API algorithm, Particle swarm optimization,
doi:10.5019/j.ijcir.2004.142
Full Text PDF
^ TOP