When, How Long and How Much? Interpretable Neural Networks for Time Series Regression by Learning to Mask and Aggregate

2025-12-03

AI Summary

In time-series regression (TSR), existing models achieve high predictive accuracy but lack interpretability; post-hoc attribution methods are coarse and unstable; and current intrinsically interpretable approaches suffer from reliance on concept-level supervision, inability to model feature interactions, limited expressiveness, and poor scalability to high-dimensional multivariate settings. This paper proposes MAGNETS: an end-to-end trainable, intrinsically interpretable neural architecture that jointly learns temporal masks and feature-combination masks to automatically discover salient time segments and meaningful variable interaction patterns. Without requiring concept annotations, MAGNETS generates human-understandable additive explanations that explicitly characterize *when*, *which features*, *for how long*, and *to what extent* each pattern influences the prediction. Evaluated on multivariate time-series regression tasks, MAGNETS achieves state-of-the-art predictive accuracy while delivering stable, fine-grained, and empirically verifiable interpretations—significantly outperforming both post-hoc explanation methods and existing intrinsically interpretable baselines.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Computer Vision: Interpretability, Explainability, and Transparency
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Time series extrinsic regression (TSER) refers to the task of predicting a continuous target variable from an input time series. It appears in many domains, including healthcare, finance, environmental monitoring, and engineering. In these settings, accurate predictions and trustworthy reasoning are both essential. Although state-of-the-art TSER models achieve strong predictive performance, they typically operate as black boxes, making it difficult to understand which temporal patterns drive their decisions. Post-hoc interpretability techniques, such as feature attribution, aim to to explain how the model arrives at its predictions, but often produce coarse, noisy, or unstable explanations. Recently, inherently interpretable approaches based on concepts, additive decompositions, or symbolic regression, have emerged as promising alternatives. However, these approaches remain limited: they require explicit supervision on the concepts themselves, often cannot capture interactions between time-series features, lack expressiveness for complex temporal patterns, and struggle to scale to high-dimensional multivariate data. To address these limitations, we propose MAGNETS (Mask-and-AGgregate NEtwork for Time Series), an inherently interpretable neural architecture for TSER. MAGNETS learns a compact set of human-understandable concepts without requiring any annotations. Each concept corresponds to a learned, mask-based aggregation over selected input features, explicitly revealing both which features drive predictions and when they matter in the sequence. Predictions are formed as combinations of these learned concepts through a transparent, additive structure, enabling clear insight into the model's decision process.

Problem

• Develop interpretable neural networks for time series regression
• Address black-box limitations in existing TSER models
• Reveal key temporal features and their timing in predictions

Innovation

• Learns mask-based concepts without annotations
• Reveals feature importance and temporal relevance
• Uses transparent additive structure for decisions