Time-series attribution maps with regularized contrastive learning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Schneider, Steffen, Laiz, Rodrigo González, Filippova, Anastasiia, Frey, Markus, Mathis, Mackenzie Weygandt
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910834788139008
author Schneider, Steffen
Laiz, Rodrigo González
Filippova, Anastasiia
Frey, Markus
Mathis, Mackenzie Weygandt
author_facet Schneider, Steffen
Laiz, Rodrigo González
Filippova, Anastasiia
Frey, Markus
Mathis, Mackenzie Weygandt
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2502_12977
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Time-series attribution maps with regularized contrastive learning
Schneider, Steffen
Laiz, Rodrigo González
Filippova, Anastasiia
Frey, Markus
Mathis, Mackenzie Weygandt
Machine Learning
Artificial Intelligence
Neurons and Cognition
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.
title Time-series attribution maps with regularized contrastive learning
topic Machine Learning
Artificial Intelligence
Neurons and Cognition
url https://arxiv.org/abs/2502.12977