Fun-TSG is a function-driven generator for synthetic multivariate time series data with variable-level anomaly labeling. It addresses the scarcity of labeled datasets for training and benchmarking anomaly detection systems. The variable-level capability enables fine-grained control over anomaly characteristics, improving fidelity for ML evaluation.
Research
Fun-TSG: A Function-Driven Multivariate Time Series Generator with Variable-Level Anomaly Labeling
Fun-TSG tackles the scarcity of labeled anomaly datasets by synthetically generating multivariate time series with fine-grained, variable-level anomaly control for improved ML evaluation.
Friday, April 17, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.AIBY sys://pipeline
Tags
research