Università Roma Tre

The course is organized into 6 modules, each with a specific focus.

The first module, Foundations, introduces the fundamental concepts of artificial intelligence and electromagnetic theory integration. It provides students with a shared vocabulary and methodological grounding in both fields to facilitate interdisciplinary understanding.

The second module, Supervised Learning for EM Modeling, explores how labeled data from electromagnetic simulations or experiments can be used to train predictive models. Emphasis is placed on regression tasks, neural networks, and data-driven approximation of field distributions and device responses.

In the third module, students examine Physics-Informed Neural Networks (PINNs) and selected unsupervised learning techniques. This module emphasizes the embedding of Maxwell’s equations directly into the learning process to ensure physically consistent predictions, especially in situations with limited training data.

The fourth module, Surrogate Modeling and Model Order Reduction, addresses strategies to replace computationally expensive full-wave simulations with fast, data-driven models. Students learn how to build compact and efficient surrogates that preserve physical fidelity while significantly accelerating design processes.

The fifth module focuses on Design Optimization, highlighting how AI techniques can reconstruct material properties, geometries, or source distributions from measurement or simulated data. It also covers constrained optimization in high-dimensional spaces using AI-enhanced solvers.

The final module, Advanced Applications and Integration with EM Software, demonstrates how to embed AI models into industrial electromagnetic workflows using tools like CST Studio Suite, HFSS, and MATLAB. Students apply their skills to real-world challenges such as antenna synthesis, metamaterial characterization, and intelligent electromagnetic environment modeling.

Main textbooks:
S. Haykin, Neural Networks and Learning Machines, 3rd ed., Pearson

C. M. Bishop, Pattern Recognition and Machine Learning, Springer

S. M. Rao, Time Domain Electromagnetics, Academic Press

Additional resources:
Goodfellow, Bengio, Courville, Deep Learning, MIT Press

Krizhevsky et al., ImageNet Classification with Deep Convolutional Neural Networks, NIPS

Selected articles from international scientific journals, including:

IEEE Trans. on Antennas and Propagation

IEEE Trans. on Microwave Theory and Techniques

IEEE Trans. on Neural Networks and Learning Systems

Nature Machine Intelligence

EPJ Applied Metamaterials

Other materials:
Lecture notes by the instructor: custom teaching materials will be provided throughout the course, including code notebooks, conceptual diagrams, datasets, and problem sets, designed to support learning and facilitate practical applications.