Esther Rodrigo Bonet is a Post-doc Researcher at ETRO, Vrije Universiteit Brussel (VUB) and ISP, Universitat de València, and an imec fellow. Her doctoral research, funded by an FWO PhD Fellowship, explored explainable physics-guided deep learning for air pollution modelling — developing architectures such as physics-guided variational graph autoencoders, deep equilibrium networks, and topology-aware explainability methods for graph neural networks. She currently extends these techniques to computational biology, health AI, and uncertainty quantification within Horizon Europe projects, supervised by Nikos Deligiannis.
Research Areas
- Explainable AI & Graph Neural Networks — Topology-aware explainability methods for graph neural networks, with applications in environmental modelling and beyond.
- Physics-Guided Deep Learning — Integrating domain-specific physical constraints into deep learning architectures such as variational graph autoencoders and deep equilibrium networks.
- Environmental & Air Quality Modelling — Graph-based deep learning for air pollution inference, forecasting, and interpretable urban environment monitoring.
- Computational Biology & Health AI — Extending explainability and uncertainty quantification to computational biology, protein interactions, and health-related AI within Horizon Europe projects.
Publications
- GF-LRP: A Method for Explaining Predictions Made by Variational Graph Auto-Encoders — IEEE Trans. Emerg. Topics Comput. Intell., 2025
- Explaining Graph Neural Networks with Topology-Aware Node Selection: Application in Air Quality Inference — IEEE Trans. Signal Inf. Process. over Netw., 2022
- Physics-Guided Graph Convolutional Deep Equilibrium Network for Environmental Data — Proc. 32nd European Signal Process. Conf. (EUSIPCO '24), 2024
- Physics-Guided Variational Graph Autoencoder for Air Quality Inference — Proc. IEEE Int. Conf. Acoust. Speech Signal Process. (ICASSP '24), 2024
- Conditional Variational Graph Autoencoder for Air Pollution Forecasting — Proc. 30th European Signal Process. Conf. (EUSIPCO '22), 2022
- Temporal Collaborative Filtering with Graph Convolutional Neural Networks — Proc. 25th Int. Conf. Pattern Recognit. (ICPR '20), 2020
News
- 2024 — Best Student Paper Award at EUSIPCO 2024 for "Physics-Guided Graph Convolutional Deep Equilibrium Networks for Environmental Data".
- 2025 — Joined ENACT Horizon Europe project on explainable AI for health applications.
- 2020 — Awarded FWO PhD Fellowship Strategic Basic Research for explainable physics-guided deep learning.