Machine Learning for Human Data is an advanced course on machine learning and neural network models with selected applications to human data (e.g., ECG, movement data, audio, biomedical images). The material will revolve around unsupervised and supervised learning, as follows. Unsupervised learning: Self Organising Maps, Neural Gas Networks, the DBSCAN algorithm, denoising auto encoders (CNN and RNN-based). Supervised learning: recurrent neural networks (RNN), graph neural networks (GNN), attention mechanisms (the transformer model), spiking neural networks (towards neuromorphic computing). A key aspect of the course is the rich laboratory activity, where students will learn to implement their learning architectures (in the Python programming language, using TensorFlow), train and test them on selected datasets.
- Teacher: Michele Rossi