Improving Sample Average Approximation Using Distributional Robustness

Sample average approximation is a popular approach to solving stochastic optimization problems. It has been widely observed that some form of robustification these problems often improves the out-of-sample performance solution estimators. In estimation problems, this improvement boils down trade-off between opposing effects bias and shrinkage. This paper aims characterize features more general exhibit behaviour when distributionally robust version sample problem used. The restricts attention quadratic for which solutions are unbiased shows expected can be calculated small amounts depends on type model used properties underlying ground-truth probability distribution random variables. was written as part New Zealand funded research project aimed improve methods in electric power industry. authors have worked together domain past 25 years.