AI Insight
L2GTX is a new model-agnostic framework that generates global explanations for time series classification by aggregating local explanations from representative instances. The method identifies temporal patterns like trends and extrema, clusters them based on importance, and selects key instances that maximize coverage of influential patterns to create concise class-wise explanations. Testing on six benchmark datasets demonstrates that L2GTX produces interpretable global explanations while maintaining consistent faithfulness to the underlying model's behavior.
Why it matters
This approach addresses a critical gap in explainable AI for time series data, enabling users to understand why deep learning models make specific classification decisions across entire classes rather than individual predictions. The model-agnostic nature makes it applicable across different architectures and domains like healthcare monitoring, financial forecasting, and industrial sensor analysis where understanding temporal decision patterns is essential.
Understand the Science
arXiv:2603.13065v2 Announce Type: replace-cross
Abstract: Deep learning models achieve high accuracy in time series classification, yet understanding their class-level decision behaviour remains challenging. Explanations for time series must respect temporal dependencies and identify patterns that recur across instances. Existing approaches face three limitations: model-agnostic XAI methods developed for images and tabular data do not readily extend to time series, global explanation synthesis for time series remains underexplored, and most existing global approaches are model-specific. We propose L2GTX, a model-agnostic framework that generates class-wise global explanations by aggregating local explanations from a representative set of instances. L2GTX extracts clusters of parameterised temporal event primitives, such as increasing or decreasing trends and local extrema, together with their importance scores from instance-level explanations produced by LOMATCE. These clusters are merged across instances to reduce redundancy, and an instance-cluster importance matrix is used to estimate global relevance. Under a user-defined instance selection budget, L2GTX selects representative instances that maximise coverage of influential clusters. Events from the selected instances are then aggregated into concise class-wise global explanations. Experiments on six benchmark time series datasets show that L2GTX produces compact and interpretable global explanations while maintaining stable global faithfulness measured as mean local surrogate fidelity.
Source: L2GTX: From Local to Global Time Series Explanations