Sequential Change Point Detection via Denoising Score Matching

2025-01-22

AI Summary

This paper addresses the problem of real-time online detection of distributional shifts in high-dimensional, complex data streams under a parameter-free assumption. We propose a novel score-based CUSUM method grounded in denoising score matching: by injecting controlled noise and estimating the data’s score function via denoising score matching, we construct an incrementally updatable nonparametric test statistic. To our knowledge, this is the first work to integrate denoising score matching into sequential change-point detection. We theoretically establish that tuning the noise scale improves detection power, and the framework supports both offline pretraining and online incremental learning. Evaluated on two synthetic benchmarks and a real-world earthquake precursor monitoring task, our method reduces average detection delay by 37% and false alarm rate by 52% compared to conventional approaches—particularly excelling in low signal-to-noise ratio and high-dimensional nonstationary regimes.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Data Stream Mining
• Search and Optimization: Distributed Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Sequential change-point detection plays a critical role in numerous real-world applications, where timely identification of distributional shifts can greatly mitigate adverse outcomes. Classical methods commonly rely on parametric density assumptions of pre- and post-change distributions, limiting their effectiveness for high-dimensional, complex data streams. This paper proposes a score-based CUSUM change-point detection, in which the score functions of the data distribution are estimated by injecting noise and applying denoising score matching. We consider both offline and online versions of score estimation. Through theoretical analysis, we demonstrate that denoising score matching can enhance detection power by effectively controlling the injected noise scale. Finally, we validate the practical efficacy of our method through numerical experiments on two synthetic datasets and a real-world earthquake precursor detection task, demonstrating its effectiveness in challenging scenarios.

Problem

• Change Detection
• Complex Data
• Non-parametric Methods

Innovation

• Score-based CUSUM
• Change-point Detection
• Robustness to Noise