Time-series attribution maps with regularized contrastive learning

2025-02-17

AI Summary

This work addresses the fundamental lack of identifiability guarantees in gradient-based attribution methods for time-series models. We propose xCEBRA, an identifiable attribution framework that integrates regularized contrastive learning with a novel attribution method—Inverted Neuron Gradient (ING)—enabling, for the first time, identifiable inference of the Jacobian matrix of the underlying data-generating process. We theoretically establish Jacobian identifiability under mild assumptions, overcoming the inherent non-identifiability limitation of existing gradient-based attribution approaches. Empirical evaluation on synthetic benchmarks demonstrates that xCEBRA accurately recovers the zero/non-zero structure of ground-truth attribution maps, significantly outperforming feature occlusion, Shapley value, and all tested gradient-based methods. Thus, xCEBRA establishes the first attribution paradigm for time-series explainable AI with formal identifiability guarantees.

Technology Category

• Machine Learning: Causal Learning
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Reasoning under Uncertainty: Causality

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Gradient-based attribution methods aim to explain decisions of deep learning models but so far lack identifiability guarantees. Here, we propose a method to generate attribution maps with identifiability guarantees by developing a regularized contrastive learning algorithm trained on time-series data plus a new attribution method called Inverted Neuron Gradient (collectively named xCEBRA). We show theoretically that xCEBRA has favorable properties for identifying the Jacobian matrix of the data generating process. Empirically, we demonstrate robust approximation of zero vs. non-zero entries in the ground-truth attribution map on synthetic datasets, and significant improvements across previous attribution methods based on feature ablation, Shapley values, and other gradient-based methods. Our work constitutes a first example of identifiable inference of time-series attribution maps and opens avenues to a better understanding of time-series data, such as for neural dynamics and decision-processes within neural networks.

Problem

• Identifiable time-series attribution maps
• Regularized contrastive learning algorithm
• Inverted Neuron Gradient method

Innovation

• Regularized contrastive learning algorithm
• Inverted Neuron Gradient method
• Identifiable time-series attribution maps