2019/01/18 by Wei-Cheng Chang, Chun-Liang Li, Chang, Wei-Cheng +6 · 4 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · Mathematics · #Algorithm #Artificial intelligence #Benchmark (surveying) #Computer science #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Generative grammar #Generative model #Kernel (algebra) #Kernel embedding of distributions #Kernel method #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine learning #Mathematics #Metabolomics and Mass Spectrometry Studies #Multiple kernel learning #Parametric statistics #Radial basis function kernel #Selection (genetic algorithm) #Statistics #Support vector machine #Time Series Analysis and Forecasting #Tree kernel #Variable kernel density estimation #cs.LG #stat.ML
paper · pdf · doi:10.48550/arxiv.1901.06077
To appear in ICLR 2019
arxiv created 2019/01/18 · openalex publication_date 2019/01/18 · arxiv updated 2019/01/21 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06
Detecting the emergence of abrupt property changes in time series is a challenging problem. Kernel two-sample test has been studied for this task which makes fewer assumptions on the distributions than traditional parametric approaches. However, selecting kernels is non-trivial in practice. Although kernel selection for two-sample test has been studied, the insufficient samples in change point detection problem hinder the success of those developed kernel selection algorithms. In this paper, we propose KL-CPD, a novel kernel learning framework for time series CPD that optimizes a lower bound of test power via an auxiliary generative model. With deep kernel parameterization, KL-CPD endows kernel two-sample test with the data-driven kernel to detect different types of change-points in real-world applications. The proposed approach significantly outperformed other state-of-the-art methods in our comparative evaluation of benchmark datasets and simulation studies.