STICERD Econometrics Seminar Series
Universal Prediction Band, Semi-Definite Programming and Variance Interpolation
Tengyuan Liang (Chicago)
Thursday 03 June 2021 15:00 - 16:30
Due to the onging coronavirus outbreak, many of our seminars and public events this year will continue as online seminars. Please check our website listings and Twitter feed @STICERD_LSE for updates.
About this event
A frequent criticism from the statistics community to modern machine learning is the lack of rigorous uncertainty quantification. Instead, the machine learning community would argue that conventional uncertainty quantification based on idealized distributional assumptions may be too restrictive for real data. Nevertheless, without a doubt, uncertainty quantification for predictive modeling is essential to statistics, learning theory, and econometrics. This paper will resolve the above inference dilemma by introducing a new method with provable uncertainty quantification via semi-definite programming. We propose a computationally efficient method to construct nonparametric, heteroskedastic prediction bands for uncertainty quantification, with or without any user-specified predictive model. The data-adaptive prediction band is universally applicable with minimal distributional assumptions, with strong non-asymptotic coverage properties, and easy to implement using standard convex programs. Our approach can be viewed as a novel variance interpolation with confidence and further leverages techniques from semi-definite programming and sum-of-squares optimization. Theoretical and numerical performances for the proposed approach for uncertainty quantification are analyzed.
STICERD Econometrics seminars are held on Thursdays in term time at 14.00-15.30, ONLINE, unless specified otherwise.
For further information please contact Lubala Chibwe, either by email: email@example.com.
Please use this link to subscribe or unsubscribe to STICERD Econometrics mailing list (emetrics).