PhD programmes - Science and Technology
Home > Machine Learning & AI Methods - Theory, Techniques, and Advanced Engineering Applications

Machine Learning & AI Methods - Theory, Techniques, and Advanced Engineering Applications

Lecturers: Marco Broccardo (UNITN/DICAM); Marco Salucci (UNITN/DICAM)

Timetableto be defined

Course Description:
Understanding and solving complex problems in the physical world has been an intelligent endeavor
of humankind. Moreover, the study of artificial intelligence (AI) embodies the dream of designing
machines like humans. Research in machine learning (ML) and, more recently, on deep learning (DL)
techniques has attracted much attention in many engineering fields. With the spreading of such
techniques, improvement in learning capacity may allow machines to “learn” from a large amount of
physical data and “master” the physical laws in certain controlled boundary conditions. In the long
run, a hybridization of fundamental physical principles with ”knowledge” from big data could unleash
numerous engineering applications that used to be impossible due to the limit of data information
and computational capabilities.
The course aims at providing a solid background knowledge on AI and ML, with a focus on recent and
competitive methodologies for the efficient and robust solution of both classification and regression
problems in advanced engineering applications. Applicative examples including exercises will corroborate
the theoretical concepts.

Course Topics:
- Basics, fundamental theory, and pillar concepts of AI and ML;
- ML as a "three-steps" learning process for regression and classification purposes;
- Space reduction via feature selection strategies: Sequential Feature Selection (SFS),
  Sequential Forward Feature Selection (SFFS);
- Space reduction via feature extraction strategies: Principal Component Analysis (PCA),
  Partial Least Squares (PLS), Sammon Mapping (SM);
- Space exploration strategies via single-shot and adaptive sampling strategies: uniform
  grid/random sampling, Latin Hypercube Sampling (LHS), LOLA-Voronoi, Output Space Filling (OSF);
- ML for classification: Support Vector Machines (SVMs), Neural Networks (NNs);
- ML for regression: Gaussian Processes (GPs), Radial Basis Function Networks (RBFN), Support
  Vector Regression (SVR);
- Basics of DL: Deep Convolutional Neural Networks (CNNs);
- Applicative examples including exercises regarding specific engineering applications of
  AI and ML.

[1] A. I. J. Forrester, A. Sobester, and A. J. Keane, Engineering Design via Surrogate Modelling:
    A Practical Guide. Hoboken, N.J.: John Wiley & Sons, 2008.
[2] A. Massa, G. Oliveri, M. Salucci, N. Anselmi, and P. Rocca, “Learning-by-examples techniques
    as applied to electromagnetics,” J. Electromagn. Waves Appl., vol. 32, no. 4, pp. 516-541, 2018
[3] M. Li, R. Guo, K. Zhang, Z. Lin, F. Yang, S. Xu, X. Chen, and A. Massa, “Machine learning in  
    electromagnetics with applications to biomedical imaging - A review,” IEEE Antennas Propag. Mag.,
    vol. 63, no. 3, pp. 39-51, Jun. 2021

Duration: 32 hours ( 4 credits)

Registration: in order to access the course please send an email to dicamphd [at]