Optimal Information Retention for Time-Series Explanations
Published in ICML 2025, 2025
We propose the Optimal Information Retention Principle for time-series explanations, where conditional mutual information defines minimizing redundancy and maximizing completeness as optimization objectives. We introduce the ORTE framework, learning a binary mask to eliminate redundant information while mining temporal patterns of explanations.
Recommended citation: Jinghang Yue, Jing Wang, Lu Zhang, Shuo Zhang, Da Li, Zhaoyang Ma, Youfang Lin. (2025). "Optimal Information Retention for Time-Series Explanations." ICML 2025.
Download Paper