A technique that finds optimal values in expensive search spaces by learning from each experiment rather than just exploring
Bayesian statistics are a different approach to data science and machine learning. One where unknown parameters are not just fixed unknown values, but random variables instead.