Self-Correcting Bayesian Optimization through Bayesian Active Learning