Estratégias Evolutivas para o Problema de Desconvolução Sı́smica

The problem of seismic deconvolution, based on the sparse hypothesis of the reflection profile, requires the optimiza- tion of a non-linear and multimodal function. In this work, we investigate the use of different evolutionary algorithms to perform the search of deconvolution filter parameters that minimize a normalized pseudo-Huber function, which have mechanisms to escape from optimal locations and explore more broadly the space of candidate solutions. The results obtained with the synthetic data show the performance progress of the deconvolution filters designed with these algorithms when compared to a stochastic gradient type method.