Reproduce “TFEC: Multivariate Time-Series Clustering via Temporal-Frequency Enhanced Contrastive Learning” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zexi Tan, Tao Xie, Haoyi Xiao, Baoyao Yang, Yuzhu Ji, An Zeng, Xiang Zhang, Yiqun Zhang

TFEC introduces a framework for multivariate time-series clustering that combines a temporal-frequency co-enhancement mechanism with a pseudo-label guided contrastive learning strategy. This approach generates low-distortion, cluster-friendly representations and achieves superior performance on various real-world datasets.

Key insights

  • Combining temporal integrity preservation (aligned cropping) with adaptive frequency-domain enrichment from semantically proximate neighbors effectively generates robust and discriminative representations without harmful distortions.
  • Integrating cluster-aware information, through pseudo-labels and high-confidence sampling, directly into contrastive learning significantly enhances the formation of distinct and well-separated clusters.
  • A dual-path architecture that jointly optimizes for representation stability and cluster distribution fidelity leads to a more robust and effective deep clustering framework for multivariate time series.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.