学术报告

Searching for Interactions in Linear Time

主 题  Searching for Interactions in Linear Time
报告人  Feng Ruan, Department of EECS at Berkeley
时 间 2021年3月30日上午, 10:00--11:00
地 点 线上会议(Zoom会议号:651 2635 2283   密码:214152)
摘 要

  We tackle the problem of variable selection with a focus on discovering interactions between variables. With p variables, there are O(p^k) possible interactions of order k making exhaustive search infeasible. We show that it is nonetheless possible to identify the variables involved in interactions (of any order) with only linear computation cost, O(p), and in a nonparametric fashion. Our algorithm is based on minimizing a non-convex objective, carefully designed to have a favorable landscape. We provide finite sample guarantees on both false positives (we show all stationary points of the objective exclude noise variables) and false negatives (we characterize the sample sizes needed for gradient descent to converge to a "good’’ stationary point).