Learning Representations of Event Time Series with Sparse Autoencoders for Anomaly Detection, Similarity Search, and Unsupervised Classification

2025-07-15

AI Summary

Irregular temporal structures and domain-specific characteristics of event time series hinder effective modeling by conventional methods. This paper proposes a multidimensional tensor representation framework for discrete event sequences: it introduces the first two- and three-dimensional tensor structures that jointly encode temporal dynamics and spectral features, and designs a sparse autoencoder to learn physically meaningful, interpretable latent representations. The framework unifies support for multiple unsupervised downstream tasks—including anomaly detection, similarity retrieval, and semantic clustering. Evaluated on real X-ray astronomical data, the model accurately separates transient source classes, significantly enhancing time-frequency pattern capture as well as anomaly detection and similarity search performance (F1 score improved by 12.6%, mean average precision for retrieval increased by 9.3%). This work establishes a generalizable and interpretable paradigm for analyzing irregular event sequences.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Event time series are sequences of discrete events occurring at irregular time intervals, each associated with a domain-specific observational modality. They are common in domains such as high-energy astrophysics, computational social science, cybersecurity, finance, healthcare, neuroscience, and seismology. Their unstructured and irregular structure poses significant challenges for extracting meaningful patterns and identifying salient phenomena using conventional techniques. We propose novel two- and three-dimensional tensor representations for event time series, coupled with sparse autoencoders that learn physically meaningful latent representations. These embeddings support a variety of downstream tasks, including anomaly detection, similarity-based retrieval, semantic clustering, and unsupervised classification. We demonstrate our approach on a real-world dataset from X-ray astronomy, showing that these representations successfully capture temporal and spectral signatures and isolate diverse classes of X-ray transients. Our framework offers a flexible, scalable, and generalizable solution for analyzing complex, irregular event time series across scientific and industrial domains.

Problem

• Representing irregular event time series for pattern extraction
• Detecting anomalies and similarities in sparse event data
• Classifying events unsupervisedly across diverse scientific domains

Innovation

• Novel tensor representations for event time series
• Sparse autoencoders for meaningful latent embeddings
• Flexible framework for diverse downstream tasks