Résumé :
This talk delves into the practical and theoretical aspects of Physics-Informed Neural Networks (PINNs) in predictive modeling, emphasizing their capability in approximating solutions to partial differential equations across various fields. Initially, we will explore key theoretical features of PINNs including learning theory, and error optimization, crucial for their effective application. Subsequently, we will showcase specific applications, like addressing the soil microbiota growth issue, illustrating the utility of PINNs in complex numerical simulations. This presentation aims to equip attendees with the essential knowledge for employing PINNs effectively in their respective research or industry projects, catering particularly to machine learning experts keen on expanding their expertise in this domain.
Bio :
Vincenzo Schiano Di Cola is an accomplished R&D Data Scientist specializing in machine learning and numerical modeling, holding a Ph.D. in Data Science from the University of Naples Federico II. His notable contributions include pioneering predictive analytics and crafting physics-informed neural networks (PINNs). His diverse research encompasses deep fake attack assessment for security, modeling proof of ID systems for safety enhancement, and analyzing museum visitors' paths clustering. Vincenzo has developed knowledge graphs across multiple domains like health prescriptions and IoT data in cultural heritage, alongside venturing into image processing and diffusion equation simulation. Besides his extensive research, he has educated high school students in computer science, mathematics, and physics, showcasing his enthusiasm for knowledge dissemination. Vincenzo’s work is well-recognized in the academic community with publications in reputable journals and presentations at various conferences, continuously seeking novel challenges to leverage his expertise in machine learning and data science.
