Reproduce “A Reliable Framework for Human-in-the-Loop Anomaly Detection in Time Series” 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
Ziquan Deng, Xiwei Xuan, Kwan-Liu Ma, Zhaodan Kong
Researchers at the University of California, Davis developed HILAD, an interactive visual analytics framework for human-in-the-loop anomaly detection in time series data. This framework allows domain experts to identify and correct systematic model biases, demonstrating improved classification accuracy and attention correctness with minimal human annotation.
Key insights
- Combining time series context similarity with model attribution similarity is crucial for effective clustering that reveals systematic model behaviors and biases.
- Scalable annotation and label propagation, guided by attribution-aware clustering, enables efficient identification and correction of model biases with minimal human effort.
- Providing users with mechanisms for corrective feedback and model refinement, beyond just explanations, is essential for truly enhancing the reliability and trustworthiness of AI systems in critical applications.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.