Unsupervised Ensemble Regression

Omer Dror, Boaz Nadler, Erhan Bilal, Yuval Kluger

Research output: Contribution to journalArticle


Consider a regression problem where there is no labeled data and the only observations are the predictions fi(xj) of m experts fi over many samples xj. With no knowledge on the accuracy of the experts, is it still possible to accurately estimate the unknown responses yj? Can one still detect the least or most accurate experts? In this work we propose a framework to study these questions, based on the assumption that the m experts have uncorrelated deviations from the optimal predictor. Assuming the first two moments of the response are known, we develop methods to detect the best and worst regressors, and derive U-PCR, a novel principal components approach for unsupervised ensemble regression. We provide theoretical support for U-PCR and illustrate its improved accuracy over the ensemble mean and median on a variety of regression problems.
Original languageEnglish
Number of pages10
StateSubmitted - 8 Mar 2017


Dive into the research topics of 'Unsupervised Ensemble Regression'. Together they form a unique fingerprint.

Cite this