Sequential change-point detection: Computation versus statistical performance

TL;DR

This paper reviews the trade-off between computational efficiency and statistical power in sequential change-point detection, emphasizing modern high-dimensional data challenges and analyzing various algorithms from classic to recent non-parametric methods.

Contribution

It introduces a new perspective focusing on computation versus statistical performance trade-offs and surveys a range of detection procedures including recent non-parametric approaches.

Findings

Highlights the importance of balancing computation and statistical power.

Provides analysis techniques for evaluating change-point detection algorithms.

Reviews both classic and modern non-parametric detection methods.

Abstract

Change-point detection studies the problem of detecting the changes in the underlying distribution of the data stream as soon as possible after the change happens. Modern large-scale, high-dimensional, and complex streaming data call for computationally (memory) efficient sequential change-point detection algorithms that are also statistically powerful. This gives rise to a computation versus statistical power trade-off, an aspect less emphasized in the past in classic literature. This tutorial takes this new perspective and reviews several sequential change-point detection procedures, ranging from classic sequential change-point detection algorithms to more recent non-parametric procedures that consider computation, memory efficiency, and model robustness in the algorithm design. Our survey also contains classic performance analysis, which still provides useful techniques for analyzing new procedures.