Joint approach of particle swarm optimization and Gibbs sampler for improving non-linear solution

We used the particle swarm optimization (PSO) and Gibbs sampler for inversion. Sampling various arrangement of the layer parameters (dim) generate M2*dim possible samples for M numbers of layer. Probability distribution function (PDF) is calculated for all models and mean model estimated using PDF with 68.27% confidence interval is the optimum. The output show better results than earlier published work. The technique used here is PSO and implemented Gibbs sampler where sampling is done with respect to distribution and no need to evaluate cost function at each point in model space. The optimized results are shown in Table1 for noise free and noisy data with true model taken from Shaw and Shalivahan (2007). The uncertainty in model parameter for noise free data are 1633.43, 1248.22, 341.02, 1878.81, 2130.98. Results show better agreement with true model.