This advanced Master-level course explores modern machine and deep learning techniques tailored for biological signal processing and human data analysis. Moving beyond static text and images, the curriculum emphasizes dynamic, spatio-temporal signals that exhibit complex correlations across both spatial dimensions (e.g., sensor arrays) and temporal domains (e.g., physiological time series).
The course places a strong, distinct focus on unsupervised learning and anomaly detection, a critical and yet often neglected skill sets, necessary to analyze unannotated biomedical data. Students will gain theoretical mastery and practical expertise across unsupervised feature extraction (SOMs, DBSCAN, autoencoders, one class support vector machines), supervised deep architectures (CNNs, RNNs, GNNs, Transformers, SNNs), and advanced model optimization.
Hands-on laboratory sessions with tested reference code prepare students to build custom end-to-end models. Assessment is project-based (groups of max 2), requiring students to analyze real-world datasets, deliver a scientific-style report, submit working code, and complete an oral defense.
The course places a strong, distinct focus on unsupervised learning and anomaly detection, a critical and yet often neglected skill sets, necessary to analyze unannotated biomedical data. Students will gain theoretical mastery and practical expertise across unsupervised feature extraction (SOMs, DBSCAN, autoencoders, one class support vector machines), supervised deep architectures (CNNs, RNNs, GNNs, Transformers, SNNs), and advanced model optimization.
Hands-on laboratory sessions with tested reference code prepare students to build custom end-to-end models. Assessment is project-based (groups of max 2), requiring students to analyze real-world datasets, deliver a scientific-style report, submit working code, and complete an oral defense.
- Docente: Michele Rossi