👋🏼 Hello there!
About Me 📜
I’m a first-year Computer Engineering PhD student in the Institute for Intelligent Networked Systems at Northeastern University, Boston. I work as a Research Assistant in Mentis Laboratory, advised by Prof. Francesco Restuccia.
I hold a Master’s degree in Artificial Intelligence & Robotics, and a Bachelor’s degree in Computer & Control Engineering, both from Sapienza University of Rome. My Research interests are Interpretability and robustness in Machine Learning, with a focus in Geometric Deep Learning.
News 📰
- METACOG 2026 A Metacognitive Framework for Test-Time Adaptation received a Top 3 Paper Award at the AAAI Fall Symposium on Artificial Metacognition, where we also present a metacognitive perspective on anomaly detection.
- NeurIPS 2026 PLACE was accepted at NeurIPS 2026, and Heterogeneous Sheaf Neural Networks at the NeurReps workshop (Proceedings Track).
- IVC 2026 Caps-ViT is published in Image and Vision Computing.
Selected Publications 📄
PLACE: Patch-Level Agnostic Concept Extraction
Advances in Neural Information Processing Systems (NeurIPS 2026)
Label-free concept extraction on the patch tokens of frozen vision transformers, with a heatmap for every concept, about 20× faster than KL-supervised extraction.
@inproceedings{onorato2026place,
title = {{PLACE}: Patch-Level Agnostic Concept Extraction},
author = {Onorato, Gabriele and Silvestri, Fabrizio and Bastian, Nathaniel D. and Restuccia, Francesco},
booktitle = {Advances in Neural Information Processing Systems},
year = {2026},
url = {https://openreview.net/forum?id=8QlhKxCYSB}
}
Caps-ViT: A Synthetic-Free Capsule Vision Transformer for Reconstruction Based Anomaly Detection and Localization
Image and Vision Computing (Elsevier, 2026)
A capsule vision transformer for industrial anomaly detection, trained only on real normal images with no synthetic defects and no external pretraining.
@article{onorato2026capsvit,
title = {{Caps-ViT}: A Synthetic-Free Capsule Vision Transformer for Reconstruction Based Anomaly Detection and Localization},
author = {Onorato, Gabriele and Tarantelli, Kristjan and Guida, Alberto and Schiavella, Claudio and Amerini, Irene},
journal = {Image and Vision Computing},
volume = {176},
pages = {106204},
year = {2026},
publisher = {Elsevier},
doi = {10.1016/j.imavis.2026.106204}
}
Heterogeneous Sheaf Neural Networks
NeurIPS 2026 Workshop on Symmetry and Geometry in Neural Representations (NeurReps, PMLR)
Cellular sheaves for heterogeneous graphs: type-aware restriction maps replace relation-specific modules, with up to 10× fewer parameters.
@inproceedings{braithwaite2026hetero,
title = {Heterogeneous Sheaf Neural Networks},
author = {Braithwaite, Luke and Borgi, Alessio and Onorato, Gabriele and Tarantelli, Kristjan and Restuccia, Francesco and Silvestri, Fabrizio and Li{\`o}, Pietro},
booktitle = {NeurIPS 2026 Workshop on Symmetry and Geometry in Neural Representations (Proceedings Track)},
series = {Proceedings of Machine Learning Research},
publisher = {PMLR},
year = {2026},
eprint = {2409.08036},
archivePrefix = {arXiv}
}
A Metacognitive Framework for Test-Time Adaptation: Taxonomy and Research Challenges
AAAI Fall Symposium on Artificial Metacognition (METACOG 2026) Top 3 Paper Award
Test-time adaptation as a two-level metacognitive system, with a taxonomy of existing protocols and the case for monitoring grounded in internal concepts.
@inproceedings{loi2026metacognitive,
title = {A Metacognitive Framework for Test-Time Adaptation: Taxonomy and Research Challenges},
author = {Loi, Dario and Onorato, Gabriele and Coviello, Antonella and Rifat, Shahriar and Bastian, Nathaniel D. and Restuccia, Francesco},
booktitle = {Proceedings of the AAAI Fall Symposium on Artificial Metacognition (METACOG)},
series = {AAAI Fall Symposium Series},
address = {Arlington, VA, USA},
month = {November},
year = {2026}
}
A Metacognitive Perspective on Anomaly Detection with Vision Foundation Models
AAAI Fall Symposium on Artificial Metacognition (METACOG 2026)
Monitoring and deciding dissociate in anomaly detection: a Mahalanobis scorer monitors better, and an Extreme Value Theory threshold recovers its segmentation without anomalous data.
@inproceedings{galiazzo2026metacognitive,
title = {A Metacognitive Perspective on Anomaly Detection with Vision Foundation Models},
author = {Galiazzo, Matteo and Onorato, Gabriele and Loi, Dario and Bastian, Nathaniel D. and Restuccia, Francesco},
booktitle = {Proceedings of the AAAI Fall Symposium on Artificial Metacognition (METACOG)},
series = {AAAI Fall Symposium Series},
address = {Arlington, VA, USA},
month = {November},
year = {2026}
}
Bayesian Optimization for Hyperparameters Tuning in Neural Networks
Bachelor's thesis, Sapienza University of Rome (arXiv 2024)
Gaussian-process Bayesian optimization with Ax and BoTorch to tune CNN hyperparameters in fewer trials.
@misc{onorato2024bayesian,
title = {Bayesian Optimization for Hyperparameters Tuning in Neural Networks},
author = {Onorato, Gabriele},
year = {2024},
eprint = {2410.21886},
archivePrefix = {arXiv},
note = {Bachelor's thesis, Sapienza University of Rome}
}My Passions ❤️
In my free time, I enjoy reading 📖📰, playing video games 👾🕹️, and padel 🎾.
