About this document

Benchmarking Deep Learning Interpretability in Time Series Predictions by Aditya Gupta is a document available to read on EtoBox.

This paper benchmarks various saliency-based interpretability methods for time series predictions across different neural architectures, revealing that these methods often fail to accurately identify feature importance over time. The authors propose a two-step Temporal Saliency Rescaling (TSR) approach to enhance the quality of saliency maps, demonstrating its effectiveness through extensive experiments. The study highlights the challenges of applying existing interpretability methods to time series data an

Author

Aditya Gupta

Language

EN