Adaptive Quickest Change Detection With Unknown Parameter

Free registration required

Executive Summary

Quickest detection of an abrupt distribution change with an unknown time varying parameter is considered. A novel adaptive approach is proposed to tackle this problem, which is shown to outperform the celebrated Parallel CUSUM Test. Performance is evaluated through theoretical analysis and numerical simulations. Quickest detection is a technique to detect distribution changes as quickly as possible based on sequential observations. It admits a wide range of applications such as quality control, medical diagnosis and intrusion detection. When both prechange and post-change distributions are completely specified, many detection procedures have been proposed under different criterias. One well-known procedure is the CUmulative Sum (CUSUM) test proposed by Page.

  • Format: PDF
  • Size: 215.34 KB