Gibbs sampling optimization in underwater sound problems

Parameter estimation in the ocean can be achieved in an optimal fashion by implementing approaches maximizing posterior probability distribution functions. Such approaches are, however, computationally intensive, often requiring the computation of complex probability distributions and searches for global maxima in spaces of a high dimension. In this work, it is shown how Gibbs Sampling, a Markov Chain Monte Carlo method, can be employed for the fast computation of posterior probability distributions, resulting in accurate and fast estimation of parameters related to problems in underwater acoustics. Source localization results obtained through maximum a posteriori estimation and optimization with Gibbs Sampling are presented and compared to results obtained with conventional methods.