Semi-supervised inference: general theory and estimation of means (Q2328051)

Estimators are proposed along with corresponding confidence intervals for the population mean. The fixed covariate procedures are introduced. Ideal semi-supervised learning and ordinary semi-supervised learning are considered where the authors analyze the asymptotic properties for both estimators. The \(l_2\)-risk upper bounds for the two proposed estimators are given. Several numerical examples are considered. This is a very nicely constructed paper with interesting statements.