题目：Surprise sampling: an optimal subsampling design

报告人： 郁文教授 （复旦大学）

Abstract :Sampling for surprise is a working principle of efficient sampling for the saving of   computational workload among other purposes. A sample is deemed ``surprising" if it has a large error of pilot prediction or a large absolute score, and will be sampled with higher sampling probability, as it in general contains more information than ``non-surprising" samples. Such sampling schemes are particularly useful when dealing with imbalanced data. Following the working principle, we propose a sampling design called surprise sampling. It caters to the specific forms of a variety of objectives. The estimation procedure is valid even if the model is misspecified and/or the  pilot estimator is inconsistent. The proposed surprise sampling includes as a special case the local case-control sampling (Fithian and Hastie, 2014), which achieves high efficiency by utilizing a clever ``adjustment" pertained only to the logistic model. The proposed estimator also performs no worse than that of Fithian and Hastie (2014) under same model specification. We present theoretical justifications of the claimed advantages and optimality of the estimation and the sampling design. Numerical studies are carried out and the evidence  in support of the theory is shown.

时间： 2019年7月5日（周五)

上午10:00-11:00

地点：565net必赢客户端本部教二楼813教室

联系人：胡涛

欢迎全体师生积极参加！