Medicine and Surgery HT

The course will explore the main paradigms of machine learning: supervised and unsupervised. Both methods considered more easily interpretable and more complex techniques will be covered. Supervised Learning (Interpretable):Logistic Regression, Decision Trees, Random Forest, XGBoost. We will analyze how these methods handle many variables and provide their significance, with a focus on Logistic Regression for binary outcomes. Outcome evaluation metrics. Unsupervised Learning (Interpretable):K-Means Clustering (with focus on cluster number input), Hierarchical Clustering (including phylogenetic trees), Lovain Clustering (for complex networks, based on communities and modularity), Network Medicine (applications from omics to clinical and social data), PCA (Principal Component Analysis for dimensionality reduction). Supervised Learning (Difficult to Interpret):Convolutional Neural Networks (CNNs), CNNs for Biomedical Images: U-NET, Semantic Image Segmentation, Transformers, Vision Transformers; From Transformers to Language Models, Examples of Language Models in Medicine.What is Explainability (XAI): riconoscere bias. Shap (SHapley Additive exPlanations). Unsupervised Learning (Difficult to Interpret):Autoencoders, Variational Autoencoders (VAE). Examples of Omics data integration and image analysis. Other Topics:A brief introduction to Deep Reinforcement Learning will be provided: what an agent is, what rewards and policies are, and an example of its application in medicine. The course will have a strong practical orientation, using real-world datasets from sources such as UK Biobank, Kaggle Datasets, and Physionet, and relying on common tools and libraries (Tensorflow and Pytorch) and program repositories (Huggingface).

Teaching Material:Lecture slides, code notebooks, and relevant scientific articles will be made available on the course's e-learning platform (Sapienza Moodle Elearning Platform).

Recommended Texts (Optional): Ian Goodfellow, Yoshua Bengio, and Aaron Courville. Deep Learning. MIT Press (free download from the official website)https://www.deeplearningbook.org/ Christopher M. Bishop, Hugh Bishop Deep Learning: Foundation and Concepts, Springer 2024 Aurelien GeronHands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems, 3rd edition.