Riccardo Finotello
Ingénieur chercheur
13
Documents
Affiliations actuelles
- 1182548
Identifiants chercheurs
- riccardo-finotello
- Arxiv : finotello_r_1
- Google Scholar : https://scholar.google.fr/citations?user=mT3o06QAAAAJ
Présentation
My research interests cover physical and computational problems, the common thread being the relation between applied mathematics and artificial intelligence, from data acquisition to the analysis. At present, I focus on two principal research areas, related to [computer vision](https://en.wikipedia.org/wiki/Computer_vision) and data science. The first is the analysis of hyperspectral images (often issued from spectroscopy techniques, such as [LIBS](https://en.wikipedia.org/wiki/Laser-induced_breakdown_spectroscopy) or [NIR](http://www.scholarpedia.org/article/Near_infrared_imaging) imagery) using supervised and unsupervised methods for object detection and (panoptic) segmentation of scenes. In particular, I am interested in geometric deep learning and [representation learning](https://paperswithcode.com/task/representation-learning) for computer vision: I study the properties of hyperspectral images using [graph neural networks](https://en.wikipedia.org/wiki/Graph_neural_network) and geometry, in order to recover the full extent of the information present in the images.The other is the application of ML and AI to experimental physics. Specifically, I took an interest in [explainable AI](https://en.wikipedia.org/wiki/Explainable_artificial_intelligence) methods and the definition of uncertainties in deep learning. I actively work in the development of deep learning techniques capable of characterize each measurement, in order to exploit statistical models in order to reject outliers and to provide a measure of the uncertainty of the prediction. I am also interested in applications of machine and deep learning to the theory of mathematics and physics, such as algebraic geometry and string theory, for their fascinating structures and their ability to provide geometrical insights on the behaviour of neural network architectures.
Domaines de recherche
Intelligence artificielle [cs.AI]
Apprentissage [cs.LG]
Physique des Hautes Energies - Théorie [hep-th]
Machine Learning [stat.ML]
Compétences
Machine Learning
Data Science
Computer Vision
Spectroscopy
Chemometrics
Geometry
Field Theory
Publications
- 4
- 4
- 4
- 3
- 1
- 1
- 2
- 2
- 1
- 1
- 1
- 1
- 1
- 5
- 3
- 3
- 1
- 1
Analyse élémentaire et isotopique par ablation laser et spectroscopie optiqueSéminaire CETAMA Analyse élémentaire et isotopique en surface/sub-surface, Oct 2023, Pau, France
Communication dans un congrès
cea-04288522v1
|
|
Machine learning for complete intersection Calabi-Yau manifoldsNeurIPS 2022 - The 36th conference on Neural Information Processing Systems, Dec 2022, New Orléans, United States. , 2022
Poster de conférence
cea-04082321v1
|
|
Deep Learning: Complete Intersection Calabi–Yau Manifolds (Chapter 5)Machine Learning in Pure Mathematics and Theoretical Physics, World Scientific, pp.151-181, 2023, 978-1-80061-369-0. ⟨10.1142/9781800613706_0005⟩
Chapitre d'ouvrage
cea-04187820v1
|
Functional renormalization group for signal detection and stochastic ergodicity breaking2023
Pré-publication, Document de travail
hal-04256551v1
|