Cargando...
Fecha
2025-11-14
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editorial
Elsevier
Citas
Resumen
The rapid advancements in machine learning have led to the deployment of complex models in critical domains such as healthcare, finance, and autonomous systems. Despite their remarkable predictive performance, the opaque nature of these models presents significant challenges for interpretability, which is essential for trust, accountability, and regulatory compliance. Explainable Artificial Intelligence (XAI) has emerged as a crucial field addressing these challenges by making black-box models more transparent. In this paper, we propose a novel model-specific local post-hoc explanation method for feedforward neural networks (FNNs) built on a solid mathematical foundation. Our approach enables the exact computation of input feature attributions for individual predictions and achieves perfect fidelity in replicating model behavior. These two properties, combined with competitive computational efficiency, demonstrate the superior performance of the proposed method compared to state-of-the-art XAI techniques. We validate the method through extensive experiments, showing its versatility across diverse types of problems. This work enhances interpretability and trust in AI systems by providing a reliable explanation framework applicable across a wide range of scenarios modeled with FNNs.
Descripción
The registered version of this article, first published in Neural Networks, is available online at the publisher's website: Elsevier, https://doi.org/10.1016/j.neunet.2025.108277
La versión registrada de este artículo, publicado por primera vez en Neural Networks, está disponible en línea en el sitio web del editor: Elsevier, https://doi.org/10.1016/j.neunet.2025.108277
La versión registrada de este artículo, publicado por primera vez en Neural Networks, está disponible en línea en el sitio web del editor: Elsevier, https://doi.org/10.1016/j.neunet.2025.108277
Categorías UNESCO
Palabras clave
Explainable AI, Feature attribution, Faithful explainability, Post-hoc explanation, Local explanation, Feedforward neural networks
Citación
Carles-Bou, J. L., & Carmona, E. J. (2026). Achieving faithful explainability in feedforward neural networks through accurately computed feature attribution. Neural Networks, 195. https://doi.org/10.1016/J.NEUNET.2025.108277
Centro
E.T.S. de Ingeniería Informática