Recovering sinusoids from noisy data using bayesian inference with simulated annealing
Mathematical and Computational Applications (2011) 16(2)
March 1994 - Present
Signal processingImage ReconstructionInverse problemsModel selectionParameter estimationBayesian Statistical inference
I am currently working on analyzing positron annihilation lifetime spectra (PALS) that involves fitting parameter-dependent model to the experimental data but, requires local nonlinear optimization routines with a reasonable guess for the search parameters. However, different sets of parameters may yield good fit for a given experimental spectrum and give also rise to ambiguities in data analysis. In order to remove these ambiguities, we develop a new procedure that combines a global nonlinear optimization routine based on Simulated Annealing algorithm (SA) with Reversible jump Markov Chain Monte-Carlo Bayesian Inference method (RJMCMC-BI) so that it provides a robust fitting tool and yields information on the reliability of the results.