|
|
Speaker:Title: Rank-Based Independence Testing in Near Linear TimeVenue: VirtualSpeaker: Chaim Even-Zohar (Alan Turing Institute, London) Title: Rank-Based Independence Testing in Near Linear Time Abstract: In 1948 Hoeffding proposed a nonparametric test that detects dependence between two continuous random variables (X,Y), based on the ranking of n paired samples (Xi,Yi). The computation of this commonly-used test statistic requires O(n log n) time. Hoeffding’s test is consistent against any dependent probability density f(x,y), but can be fooled by other bivariate distributions with continuous margins. Variants of this test with stronger consistency have been considered in works by Blum, Kiefer, and Rosenblatt, Yanagimoto, and Bergsma and Dassios, and others. The so far best known algorithms to compute them have required quadratic time. We present an algorithm that computes these improved tests… |
|
|
|
|
|
|
|
|
Speaker:Title: 2/16/2021 Computer Science for MathematiciansVenue: virtualSpeaker: Michael P. Kim (UC Berkeley) Title: Outcome Indistinguishability Abstract: Prediction algorithms assign numbers to individuals that are popularly understood as individual “probabilities” — e.g., what is the probability of 5-year survival after cancer diagnosis? — and which increasingly form the basis for life-altering decisions. The understanding of individual probabilities in the context of such unrepeatable events has been the focus of intense study for decades within probability theory, statistics, and philosophy. Building off of notions developed in complexity theory and cryptography, we introduce and study Outcome Indistinguishability (OI). OI predictors yield a model of probabilities that cannot be efficiently refuted on the basis of the real-life observations produced by Nature. We investigate a hierarchy of OI definitions, whose stringency… |
|
|
|
|
|
|
|
|
|
|