Manifold-regularised Large-Margin ℓ_p-SVDD for Multidimensional Time Series Anomaly Detection

2025/07/31 by Shervin Rahimzadeh Arashloo, Arashloo, Shervin Rahimzadeh · 1 voice
Computer Science · #Anomaly Detection Techniques and Applications #Anomaly detection #Distribution (mathematics) #Exploit #FOS: Computer and information sciences #Face and Expression Recognition #Machine Learning (cs.LG) #Manifold (fluid mechanics) #Series (stratigraphy) #Smoothness #Time Series Analysis and Forecasting #Time series #cs.LG

paper · pdf · doi:10.48550/arxiv.2507.23449

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2025/07/31 · arxiv published 2025/07/31 · arxiv updated 2025/08/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

We generalise the recently introduced large-margin ℓ_p-SVDD approach to exploit the geometry of data distribution via manifold regularising for time series anomaly detection. Specifically, we formulate a manifold-regularised variant of the ℓ_p-SVDD method to encourage label smoothness on the underlying manifold to capture structural information for improved detection performance. Drawing on an existing Representer theorem, we then provide an effective optimisation technique for the proposed method. We theoretically study the proposed approach using Rademacher complexities to analyse its generalisation performance and also provide an experimental assessment of the proposed method across various data sets to compare its performance against other methods.

Citations

• TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis
• TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data
• Accelerated Graph Learning from Smooth Signals
• Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy
• TS2Vec: Towards Universal Representation of Time Series
• Multivariate Time-series Anomaly Detection via Graph Attention Network
• A Multimodal Anomaly Detector for Robot-Assisted Feeding Using an LSTM-based Variational Autoencoder
• Kernels for sequentially ordered data

Discussions

• State-of-the-art time-series anomaly detection is still a glorified parlor trick: it tells you after the bank collapses, after the turbine explodes, after the patient dies—then pats itself on the back [bsky, 1 points, 0 comments]

Related

• Finding Pegasus: Enhancing Unsupervised Anomaly Detection in High-Dimensional Data using a Manifold-Based Approach
• Energy-Based Models for Anomaly Detection: A Manifold Diffusion Recovery Approach
• Warping Resilient Scalable Anomaly Detection in Time Series
• Robust Projection based Anomaly Extraction (RPE) in Univariate Time-Series
• Deep Convolutional Large-Margin ℓ_p-SVDD for Visual Anomaly Detection