Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a distributionally robust framework for risk-sensitive estimation of an unknown signal from observed data, leveraging the type-2 Wasserstein ball to define an ambiguity set for the joint distribution of the signals. The authors focus on minimizing the worst-case conditional value-at-risk (CVaR) of the squared estimation error, leading to the derivation of affine estimators that can be computed through a semidefinite programming approach. The proposed method is validated on a wholesale electricity price forecasting task, demonstrating a significant reduction in out-of-sample CVaR compared to traditional estimation techniques.
Affine estimators derived from a Wasserstein ambiguity set can drastically reduce forecasting errors in high-stakes environments like electricity pricing.
We propose a distributionally robust approach to risk-sensitive estimation of an unknown signal x from an observed signal y. The unknown signal and observation are modeled as random vectors whose joint probability distribution is unknown, but assumed to belong to a given type-2 Wasserstein ball of distributions, termed the ambiguity set. The performance of an estimator is measured according to the conditional value-at-risk (CVaR) of the squared estimation error. Within this framework, we study the problem of computing affine estimators that minimize the worst-case CVaR over all distributions in the given ambiguity set. As our main result, we show that, when the nominal distribution at the center of the Wasserstein ball is finitely supported, such estimators can be exactly computed by solving a tractable semidefinite program. We evaluate the proposed estimators on a wholesale electricity price forecasting task using real market data and show that they deliver lower out-of-sample CVaR of squared error compared to existing methods.