B8375 - MACHINE LEARNING PER L'ELABORAZIONE DEI SEGNALI LM

Learning outcomes

At the end of the course, students will possess the skills required to design and implement machine learning solutions in engineering contexts, with particular focus on signal processing and telecommunications. In particular, students will be able to:

• Apply machine learning models for signal compression, filtering, and prediction;
• Design and use machine learning algorithms to enhance the performance of signal processing and communication systems;
• Integrate machine learning techniques with traditional signal processing methods.
  

Course contents

• Review of signals, noise, and probabilistic models.
• Representation of multidimensional signals (time–frequency–space): Fourier transform, short-time Fourier transform (STFT), spectrogram, and an introduction to wavelet transforms.
• Digital filtering of signals.
• Linear regression and regularization techniques.
• Classification of one-dimensional and multidimensional signals.
• Model selection and evaluation: cross-validation, performance metrics, overfitting, underfitting, regularization.
• Clustering techniques: Gaussian mixture models (GMM), parameter estimation using expectation-maximization (EM).
• Dimensionality reduction: principal component analysis (PCA), kernel PCA (KPCA).
• Source separation using independent component analysis (ICA).
• Variational Bayesian factor analysis (VBFA).
• Model-driven machine learning techniques:
• The concept of hybrid learning: data + model;
• Examples of architectures inspired by physical or algorithmic models;
• Model-based neural networks (MBNN): motivations and advantages over black-box models.
• Applications: separation and classification of audio and sonar/radar sources; signal compression; extraction of latent components in multichannel signals (e.g., the cocktail party problem)

 

Readings/Bibliography

• D. J. C. MacKay, Information Theory, Inference and Learning Algorithms. Cambridge University Press, 2003.
• S. Theodoridis, Machine Learning: A Bayesian and Optimization Perspective, 2nd ed. Academic Press, 2020.
• C. M. Bishop, Pattern Recognition and Machine Learning. Springer, 2006.
• T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd ed. Springer, 2009.
• J. Watt, R. Borhani, and A. K. Katsaggelos, Machine Learning Refined: Foundations, Algorithms, and Applications, 2nd ed. Cambridge University Press, 2020.

Teaching methods

The course includes in-class lectures and in-depth seminars. Selected signal processing algorithms based on machine learning techniques will be implemented using Matlab and Python during the lectures and made available to students.
Examples covered include: separation and classification of audio and sonar/radar sources; signal compression; extraction of latent components in multichannel signals (e.g., the cocktail party problem).

Assessment methods

Student learning is assessed through a two-part examination:

• A project assignment, which may consist of the implementation of an application related to one or more topics covered in the course. Submission of the corresponding documentation (including any developed software) is required in order to access the oral examination.
• An oral examination, which includes a discussion of the project and questions on the theoretical tools introduced during the course.
  

The final grade, expressed on a 30-point scale, reflects the evaluation of both components.

The project aims to assess problem-solving skills related to the course topics, while the oral examination is intended to verify the acquisition of the theoretical knowledge outlined in the syllabus.

Teaching tools

Lecture notes, slides, exercises, and code examples will be made available on the Virtuale platform.

Office hours

See the website of Andrea Giorgetti