Environmental & Water Resource Systems
School of Civil & Environmental Engineering
I am comparing different optimization algorithms when they are applied to the Bayesian calibration program called SOARS. Usually, optimization algorithms are compared using the "best value found so far" metric. The objective of SOARS is to find not just the maximum but also the distribution of the parameters' posterior probability; therefore, this comparison instead uses a new, more applicable metric that compares the response surface created by the optimization algorithms' evaluations and the true posterior.