Syllabus: shared with students

Syllabus: materials shared with students directly

Syllabus: The course comprises five sessions of two hours each.
1. Acquiring Data and Training Models: The Limits of Lawful Development. – Web scraping and unauthorised access; copyright offences and text and data mining exceptions; unlawful data acquisition; criminal-law implications of prohibited material in training datasets.
2. Designing and Placing AI on the Market: Prohibited Practices under the EU AI Act. – Rationale, scope and application of Article 5 and the relevant Commission Guidelines. Case-based analysis of prohibited practices and their implications for AI design and development.
3. Deploying and Using AI: Third-Party Misuse and the Allocation of Criminal Liability. – Distinctions between unintended failure, foreseeable misuse, abuse and criminal use. Responsibilities of users, providers and deployers; new offences in Italian and comparative law.
4. Testing and Securing AI: Preventive Duties, Red Teaming and Vulnerability Research. Preventive duties and the actors to whom they apply. Legal implications of jailbreaking and adversarial testing; safeguards for good-faith security research.
5. Interdisciplinary Lab: Preventing AI Misuse. – Guided group work on a misuse scenario, identifying potential harms, relevant actors and feasible preventive measures.

Syllabus: Lesson 1 Introduction to Human-AI Coevolution. An overview of the course topic, delivered as a lecture with interactive discussions to engage students. (4 hours)

Lesson 2 Recommender systems: a recap. Overview of the main technologies related to recommender systems.

Lesson 3 The Human-AI Feedback Loop. Exploration and formalisation of feedback loops in ML models, with practical examples from social media, online retail, urban mapping, and chatbot platforms. (4 hours)

Lesson 4 Types of Experiments. An overview of experimental methods to evaluate feedback loop impacts, illustrated with examples from diverse human-AI ecosystems. (4 hours)

Lesson 5 Unintended Consequences in Social Media. Case studies of unintended AI recommendation effects on platforms like Facebook, Twitter, and YouTube.

Lesson 6 Unintended Consequences in Online Retail. Real-world examples of AI recommendation impacts on platforms like Amazon and Spotify.

Lesson 7 Unintended Consequences in Urban Mapping. Analysis of unintended effects of AI recommendations on platforms such as Google Maps and Airbnb.

Lesson 8 Unintended Consequences in Chatbots. Examples of challenges posed by AI-generated chatbots like LLAMA and ChatGPT.

Lesson 9 Open Challenges in Human-AI Coevolution. A discussion of unresolved technical, legal, and political issues in measuring human-AI coevolution.

Lesson 10 DSA Art. 40: a guide for data access requests

Syllabus: The course will be organised in 10 2-hours lectures:

Lecture 1 Foundations and taxonomy
Motivations and definitions for hybrid intelligence and hybrid decision-making systems; automation and augmentation; complementary team performance; agents, roles, timing, and information flows; taxonomy of human oversight, learning to abstain, learning to defer, and learning together.

Lecture 2 Human judgment and decision theory
Decisions under uncertainty; states, actions, losses, expected utility, and Bayes decision rules; risk attitudes; heuristics and biases; algorithm aversion and appreciation; advice taking and the weight-of-advice measure.

Lecture 3 Human oversight explanations and governance
Human-in-the-loop, human-on-the-loop, and human-in-command patterns; meaningful oversight; automation bias; appropriate reliance; explanations as decision support; oversight requirements for high-risk AI systems; failure modes in nominal human review.

Lecture 4 Selective classification
Reject-option classification; predictor and selection function; coverage and selective risk; cost-based and coverage-constrained formulations; ambiguity and novelty rejection; post-hoc and integrated rejectors; SelectiveNet and the role of its auxiliary objective.

Lecture 5 Practical selective prediction
Confidence and rejection scores based on maximum probability, margin, entropy, ensembles, and learned confidence estimators; ranking versus probability calibration; area under the risk-coverage curve; coverage at a target risk; validation, subgroup checks, and behavior under distribution shift.

Lecture 6 Learning to defer
Definition of deferral systems, system risk with a human component; data requirements for observing human responses; Bayes and Fisher consistency; realizable consistency; excess risk and estimation error; hypothesis-set-specific consistency bounds (H-consistency bounds); minimizability gaps. two-stage and joint training; consistent surrogate losses; softmax and one-vs-all formulations.

Lecture 7 Practical Learning to Defer
Practical aspects of learning to defer, practical implementation of deferral losses in python. Evaluation of deferral systems, budgets and performance curves. Practical two-stage and joint training; class imbalance problems and classification reports for deferral systems.

Lecture 8 Learning together
Active learning and query design; learning using privileged information; distillation from human knowledge; labels, preferences, rationales, feature feedback, and explanation corrections; explanatory interactive learning; noisy or strategic feedback; reinforcement learning with human feedback, LLM interaction.

Lecture 9 Human AI interaction and evaluation
User study design: trust versus observable reliance; over-reliance and under-reliance; mental models; cognitive forcing; human, AI, and team baselines; complementarity and oracle routing; workload, decision time, fairness, robustness, and cost; within- and between-subject user studies; offline evaluation limits and deployment monitoring.

Lecture 10 End to end integration and open problems
Integration of task definition, human and model baselines, uncertainty estimates, routing policy, interface, and evaluation; explainability of abstention systems. Deferral system with added feedback.

Syllabus: 1. Overview on the regulatory framework (EU strategy on data, GDPR, AI-Act, DSA) impacting on data-driven and ai-based research life-cycles.
2. Protocols for developers, deployers, and providers of AI-based systems to process general and sensitive data.
3. Case-studies on vulnerable users: consumers, patients, children, workers.

Syllabus: Syllabus:

� Introduction to Mechanistic Interpretability and LLMs recap
� Model Structural Properties: Outlier Dimensions (Theory and Hands on)
� Model Conditioning: Task Vectors (Theory and Hands on)
� Model Interpretation: Sparse Autoencoders (Theory and Hands on)
� Guest Lectures on Outliers and Sparse Autoencoders
� Recent Developments

Syllabus: **Detailed Syllabus**

– Introduction & Preprocessing

– Time Series, Datasets and Problems
– Missing Values and Anomalies
– Normalizations
– Time Series Components
– Stationarity

– Distances, Approximations & Global Features

– Euclidean Distance
– Dynamic Time Warping
– Piecewise Aggregate Approximation
– Symbolic Aggregate approXimation
– Discrete Fourier Transformation
– Symbolic Fourier Approximation�
– Singular Value Decomposition
– Principal Component Analysis
– Global Structural Features

– Classification & Regression�

– Instance-based Models
– Linear Models
– Tree-based Models
– Shapelet-based Models
– Dictionary-based Models
– Deep Learning-based Models
– Kernel-based Models
– Ensemble-based Models

– Forecasting

– Simple Forecasting Methods
– Exponential Smoothing
– Autoregressive Methods
– Forecasting via Reduced Regression
– Deep Learning-based Models
– Probabilistic Forecasting

Syllabus: PDAI2-PM introduces students to recent data-driven techniques where the main component is the process that generated the data (the data generating process). This is a particularly hot topic, with many companies and researchers involved (see, e.g., the list of industrial that sponsored the reference conference in 2023 https://icpmconference.org/2023/sponsor-and-exhibition/). We will consider techniques known as Process Mining, in which logs generated during the execution of a process (e.g., an industrial production process, business processes, social system �processes�) are used to infer the structure of the process. Questions of interest are, e.g.: What is the actual process being executed? Are there possibilities for improvement? Does the actual process conform to the intended reference process?

Syllabus: shared with partecipating students

Syllabus: Topics
1. Introduction
Presentation of the main topics of the course.
2. Beyond the Gaussian (theory and examples from time series in Econophysics)
Sums of independent random variables; from Bernoulli to Binomial; from Binomial to Gaussian or to Poisson; the Central Limit Theorem; random walks versus empirical log-return distributions in financial markets; alpha-stable distributions and their asymptotic properties; non-Gaussian attractors and the Generalized Central Limit Theorem; stylised facts of univariate financial time series.
3. Beyond independence (in multivariate time series and correlation matrices)
Correlation matrices (definition and examples from multivariate time series in financial markets and neuroscience); correlation networks; Random Matrix Theory, the Wishart ensemble and the Marcenko-Pastur eigenvalue distribution; interpretation of empirical eigenvalues and eigenvectors of correlation matrices; decomposition of correlation matrices in random, group, and global modes; community detection for correlation matrices.
4. Beyond ensemble equivalence
Definitions of equivalence of canonical and microcanonical ensembles in statistical physics (measure, macrostate and thermodynamic equivalence); evidences for the breakdown of ensemble equivalence in physical systems and networks; generalized correspondence between ensembles; minimum description length of canonical and microcanonical models.
5. Beyond Shannon entropy
Shannon-Khinchin and Shore-Johnson axioms; generalizations of the axioms; the Uffink family of entropies and Renyi entropy; the uninformativeness axiom and the maximum-likelihood estimation of the entropic parameter; the generalized maximum-entropy principle; the q-exponential distribution.
6. Beyond scale separation in network structure
Challenges around coarse-graining in networks; different approaches to network coarse-graining: geometric, Laplacian and multiscale network renormalization; applications of multiscale network renormalization in the quenched and annealed settings.
7. Beyond scale separation in network dynamics
Coupling dynamics and structure in network models; the Bak-Sneppen model and self-organized criticality; the fitness network model; feedback between dynamics and topology in self-organized network models; the Bouchaud-Mezard model and real-world networks.

Syllabus: Syllabus: Introduction to epidemiology; Relation and agreement; Linear regression; Logistic regression; ROC curve; Dealing with collinearity, confounding and interaction; Survival analysis;

Syllabus: Program at a glance

Cloud Computing
Basic concepts
Scalability and elasticity in cloud systems
Fault-tolerance and replication
Real-time cloud services
Operations and devops engineering
Big Data and Analytics
Basic concepts
Real-time data streaming and analytics
Distributed file-system
SQL vs NoSQL data-base systems
Big-Data and the Internet of Things
Platforms
Overview of public cloud services (AWS EC2, Google GCP, …)
Apache Hadoop, Storm, Spark
Map Reduce
OpenStack

Syllabus: Program at a glance

Virtualization Fundamentals
KVM Command-Line Interface
libvirt and virtual-manager
Virtual Switching on Linux
brctl and OpenVSwitch
Containers
LXC and netns
Public Cloud Services
AWS EC2, CloudWatch
AWS S3, DynamoDB
Open-source cloud platforms
OpenStack Nova, Glance, Neutron
OpenStack Heat/Senlin, Ceilometer/Monasca
Kubernetes
Platforms for Big Data and Analytics
Map Reduce
Apache Spark

Syllabus: See https://sites.google.com/view/mariomartinoli/teaching?authuser=0

Syllabus: Cinzia Di Dio, Module 1 � Embodied cognition
Neuropsychological background
Embodied and disembodied intellingence: what differences?Antonella Marchetti, Module 2 � Theory of Mind
Theoretical introduction
Developmental steps
Methodological issues
Measuring Theory of MindFederico Manzi, Module 3 �Early social cognition
Look at me: gaze following and social cognition
Follow me: action understanding
Mirror me: the sense of imitationLaura Miraglia, Module 4 � Affective robotics
Psycho-physiological background
Emotional resonance: what matters?
Experimental issuesGiulia Peretti, Module � Educational robotics
Theoretical background
Robots in schools: what can they do?
Experimental design with infants and toddlers

Syllabus: Lecture 1: Course Introduction � Motivation for XAI � What is an explanation � The role of �Explanation� in the novel ML process: the assessment guidelines for trustworthy AI

Lecture 2: The taxonomy of XAI methods for Machine Learning � Overview of explanation methods � Post-hoc explanation methods, Feature Attribution explanation for Tabular Data (LIME and SHAP)

Lecture 3: Post-hoc explanation methods, Rule-based explanations for Tabular Data (LORE and DR-XAI) and evaluation metrics

Lecture 4: Post hoc methods (on Images data) � IntGrad, Lime, GRADCAM(++), SmoothGRAD

Hands-on: XAI methods from XAI Library � Post hoc methods part 1 (tabular) (Lore Lime Anchor Shap)

Lecture 5: Continued Post hoc methods (on Images data) � Post hoc methods using latent space (exemplars and counterexemplars) � ABELE

Lecture 6: Contrastive Reasoning: counterfactual

Hands-on: XAI methods from XAI Library � Post hoc methods part 2 (tabular) (Lore Lime Anchor Shap)

Lecture 7: Overview of transparent by-design explanation methods � Explaining by design & Global Explainer: EBM, Tlace, TREPAN, GlocalX, Protonet, Interpretable Models

Lecture 8: Evaluation of explanations: user studies and Use Case (FAIR LEns) OR Time Series Explanations

Hands-on: Post hoc methods (Images) � IntGrad, Lime, GRADCAM(++), SmoothGRAD, part 1

Lecture 9: Trustworthy AI Pipeline OR Time Series Explanations + seminar (a team of 2/3)

Lecture 10: LLM and XAI + seminar (a team of 2/3)

Hands-on: Post hoc methods (Images) � IntGrad, Lime, GRADCAM(++), SmoothGRAD + seminar (a team of 2/3)
Seminar (a team of 2/3)

Syllabus: Introduction to graph theory. Empirical properties of complex networks (scale invariance of the degree, small-world phenomenon, modularity). Network representations (monopartite, bipartite and multilayer; binary and weighted; undirected and directed networks; unsigned and signed networks; hypergraphs; simplicial complexes). Centrality. Ranking and reputation algorithms. Mesoscale structures (communities, core-periphery and bow-tie structures). A primer on dynamical models: Watts-Strogatz and Barabasi-Albert models. A primer on static models: Erdos-Renyi, Chung-Lu and fitness models

Syllabus: Topics
1. Introduction
Presentation of the course; examples of complex systems in physics, biology, social science and economics; aspects of complexity.
2. From small to large systems
Quick overview of the three-body problem; the magnetic pendulum; emergent randomness; empirical �principles� of thermodynamics; kinetic theory; statistical thermodynamics; principles of statistical physics (ergodicity, metric transitivity, equal a-priori probability, ensembles versus trajectories); entropy in thermodynamics and physics.
3. Entropy in Information Theory and Statistical Inference
Shannon-Khinchin axioms; Shannon Entropy; Jaynes� Maximum-Entropy Principle; statistical inference; the Maximum Likelihood principle; canonical and microcanonical ensembles; entropy-likelihood relationships; canonical covariance of the constraints.
4. Ensembles of random graphs
Canonical and microcanonical graph ensembles; the Erd�s-R�nyi model; link stub reconnection; the local rewiring algorithm; the Chung-Lu model; the Park-Newman model.
5. Maximum-entropy (re)formulation of network models
Canonical ensembles with given number of links and given degree sequence (the canonical Binary Undirected Configuration Model).
6. The Configuration Model at work
Pattern detection (assortativity and clustering) using the Binary Undirected Configuration Model; graph modelling (the International Trade Network) using the Binary Undirected Configuration Model; the Binary Directed Configuration Model: derivation and use for pattern detection.
7. Network Reconstruction
The �fitness ansatz� for economic and financial networks; reconstruction of interbank and interfirm networks from aggregate input and output flows.
8. Reciprocity
Introduction to reciprocity in directed networks; definition of reciprocity indicators; joint and conditional reciprocation probabilities; (non-)reciprocated degrees; the Binary Reciprocal Configuration Model; use of the model for triadic motif detection.
9. Triadic motifs detection
Applications to food webs, international trade, supply networks, interbank networks.
10. Beyond Shannon entropy
Shannon-Khinchin vs Shore-Johnson axioms; generalizations of the axioms; the Uffink family of entropies and Renyi entropy; the uninformativeness axiom and the maximum-likelihood estimation of the entropic parameter; the generalized maximum-entropy principle; the q-exponential distribution.

Syllabus: How to build statistical models in a principled way: a review of maximum-entropy models. From null models to true models: weighted reciprocal configuration models and block-structured models. Bipartite formalism for Exponential Random Graph models. Continuous formalism for Exponential Random Graph models. Conditional framework for discrete and continuous Exponential Random Graph models. Information criteria for model selection (Likelihood Ratio Test, Akaike Information Criterion, Bayesian Information Criterion, Minimum Description Length). Applications to economic and financial systems. The course will include an overview of ongoing research carried out by Networks@IMT, thereby offering directions for possible PhD projects in this area

Syllabus: None

Syllabus: Networks growth. Erdos-Renyi random networks. Node degree distribution. Limit for large networks. Clustering coefficient. Average path length. Giant component and phase transition.
Subcritical and supercritical regimes of the Erdos-Renyi random network. Connected regime of the Erdos-Renyi random network. Watts-Strogatz model. Small worldness. Watts-Strogatz degree distribution. Spectral properties of Erdos-Renyi and Watts-Strogatz.Preferential attachment: Degree dynamics. Preferential Attachment: Degree distribution. Growth and preferential attachment. Non-linear preferential attachment. Preferential attachment with fitness. Evolving networks. Bianconi-Barabasi model and Bose-Einstein condensation. Epidemic modelling. Network epidemics: general introduction.Network epidemics: degree fluctuations and the paradox of heterogeneity. Non-linear coupling. The Kuramoto model on networks. Markovian random walk on networks.Social contagion: Cascades of information and innovations. Social influence, homophily, and the environment. Complex vs. simple contagion. Threshold and generalized contagion models (Granovetter, Watts, and Centola-Macy models).Opinion dynamics: Voter model, Majority rule model, and Sznajd model. Social impact theory. Bounded confidence models (Deffuant, Hegselmann-Krause).Cultural dynamics: Social influence and homophily in cultural evolution. Axelrod model and variants. Models of language dynamics (evolutionary and quasispecies models, the Naming Game). Language competition.Phase Transition in Complex Networks. Ferro-paramagnetic transition. The Ising model. Mean-field solution of the Ising model. Ising model on a Watts-Strogatz network. Ising model on a network with a generic degree distribution. Ising model on a scale-free network.Introduction to higher-order systems. Undirected hypergraphs. Directed hypergraphs. k-uniform hypergraphs. Bipartite representation of hypergraphs. Simplicial complexes. Nerve complex.Nerve complex. Nerve theorem. Flag complex. Delaunay triangulation. Cech complex. Vietoris-Rips complex. Filtering function and partial order. Oriented simplices. q-chains. Boundary operators. Cycles.The homology group. Betti numbers. Persistent homology.Random hypergraphs. Preferential attachment for hypergraphs. Complexes based on homogeneous random graphs. Linal-Meshulam complex. Multi-parameter complexes. Homological connectivity.

Hypergraph weighted adjacency matrix and Hypergraph Laplacian. d-dimensional dynamical systems on hypergraphs. Stability analysis. Turing patterns. Synchronisation of Stuart-Landau oscillators.

Tensorial representation of simplicial complexes. Generalised Laplacian. N-dynamical systems interacting via (d+1)-body interactions. The case of 3-body interactions. Stability analysis. Kuramoto model with 3-body interactions.

Syllabus: What is a complex system, and why can the brain be considered complex?
Neurovascular coupling.
The BOLD signal.
From BOLD time-series to a graph.
Molecular diffusion of water in the brain and fiber-tracking,
The structural architecture of the brain.
Electrophysiological signals.
Origin of EEG signal.
Origin of MEG signal.
Definition of functional networks from M/EEG time series.
Topological measures used in neuroscience and their application.
Higher-order interactions in the brain.
Models of communication in the brain.
Communication through synchronisation.
Synchronisation and symmetry.
Coarse-graining network neuroscience.
Higher-order interactions.
Null-models for Network Neuroscience.
Protein-protein interaction networks (interactome).
Human disease networks (diseasome).
Boolean networks for system biology.

Syllabus: Literature review about network reconstruction. Early attempts to infer a network structure from partial information (MaxEnt approach; the copula approach; MECAPM; Iterative Proportional Fitting algorithm; Minimum Density algorithm). Monopartite and bipartite financial networks reconstruction via the fitness model. Systemic risk estimation. How to build statistical models in a principled way: maximum-entropy models. Econometric VS maximum-entropy models: a comparison

Syllabus: The exam will consist of a project and its presentation with a discussion

1) Module 1 (2 hours): Introductory Module
Introduction to NLP & Generative AI

Module 2 (6 hours): Advanced Concepts on Generative AI
Text Generation Models (Transformer Architecture)
Transfer Learning and Computing on Large Scale
Image Generation & Text to Image Models

Module 3 (12 hours): Responsible Generative AI
A. Potential risks of using generative AI models, and a recipe for designing trustworthy-by-design generative AI applications
B. Student-led seminars each followed by an open discussion session

Module 4 (10 hours): Hands-on Sessions
A. Fine-Tuning LLMs
B. Building LLM Applications Using Retrieval-Augmented Generation (RAG)

Syllabus: – Formulating Hypotheses
– From Simple Graphs to Advanced Models
– Data Collection: Network Sampling
– Data Collection: API & Web Scraping
– Agent Based Modeling
– Graph Transformation: Backboning
– Enriching Topologies: Feature-rich modeling
– How to Validate: statistical significance of network based analyses
– Experiments from A to Z: Case studies

Syllabus: Module 1 (INF/01, 3 ECTS)
� Noise and robust statistics
� Statistical decision theory
� Classifier calibration, conformal methods and learning to reject/defer
� Bayesian inference
� Causal inference: structured causal model, potential outcome model
� PCA and embeddings in LLM.
Module 2 (SECS-S/01, 3 ECTS)
� Feature selection and regularization techniques for high-dimensional Linear and Generalized Linear Models
� Feature screening algorithms for ultra-high dimensional supervised problems
� Supervised dimension reduction; Sufficient Dimension Reduction and related techniques
� Subsampling/partitioning approaches for ultra-high sample sizes
� Under- and oversampling approaches for data rebalancing.

Syllabus: Introduction to the course and to time-to-event data.
Analysis and regression models for time-to-event data.
Regression models for time-to-event data.
Competing-risk-analysis.
Introduction to regression models for constrained data.
Beta regression.
Tobit regression.

Syllabus: 1) Introduction to bioinformatics

2) Biological databases

3) Pairwise sequence alignments

4) Basic Local Alignment Search Tool (BLAST)

5) Multiple sequence Alignment

6) Molecular Phylogenetics

7) Protein domains and proteome modularity

8) Protein structure analysis, alignments and classification

9) Protein structure prediction

10) Biomolecular interaction networks

11) hands-on (based on available freely available webservers and softwares): a) sequence alignments, b) protein structure analysis and prediction, c) molecular interaction network analysis (visualization, annotation and functional enrichment)

Syllabus: no syllabus

Syllabus: This course trains students to design, run and document computational workflows for scientific research. It starts from the Unix shell and Python, moves through the scientific Python ecosystem, software engineering practice and high-performance computing, and ends with modern machine learning and deep learning, applied to problems in particle physics and medical physics.
The course is laboratory-driven: every lecture is paired with a guided hands-on session, and the final project integrates the skills acquired across all modules.

Syllabus: Industrial and academic projects, and learning experiences, proposed by medical doctors, biologists and biotechnologists, pharmacologists, sport trainers and, more generally, operators from the med tech, bio tech, pharma tech, sport and life sciences at large. A non-exhaustive list of applications is covering collection, integration and interpretation of multi-modal and multi-omics data, biomarker identification, pathway analysis, target identification, dose-regimen optimization, clinical trials, cohort identification, drug mechanism of actions, medical imaging, platform design for the Health Data Space, etc.

The course will be organized in three main phases: the first one will include seminars and lectures concerning with potential projects/problems on biomed/sport/wellness/nutrition data to address/solve with advanced software solutions, by exploiting the techniques learned in the two-years of this master degree; the second phase will require students to form small groups, choose one of the projects to address, and implement it in the rest of the course; the third last phase will require every group to provide a pitch describing the results achieved with its own project.

Projects will come from real case studies (research centers, industry, clinical centers).

Syllabus: Module 1: Spatial and Mobility Data
Fundamentals of Geographical Information Systems
Geometrical entities (points, polygons)
Basic operations
Spatial Reference Systems
Data formats
Trajectory and Flows
Spatial TessellationsDigital spatial and mobility data
Census data and travel surveys
Mobile Phone Data
GPS data
Social media data
Other data (POIs, Road Networks, etc.)Preprocessing mobility data
filtering
compression
stop detection
trajectory segmentation
trajectory similarity and clusteringModule 2: Patterns and Laws
spatial analysis (point patterns, spatial autocorrelation, K. scan statistics)
frequent pattern analysis
individual mobility laws/patterns
collective mobility laws/patternsModule 3: Predictive and Generative Models
Prediction
Next-location prediction
Trajectory generation
Flow generation

Syllabus: 1) Introduction: the Knowledge Discovery process (6 hours)
KDD Process: all steps in a nutshell (2h)
Data understanding and Data exploration. Introduction to the data science platform KNIME (2h)
Practicals: hands-on on simple case studies using python libraries (Introduction to NumPy, Pandas and Seaborn) (2h)
2) Unsupervised learning methods and practicals (8 hours)
Clustering: basic concepts and major algorithms for centroid, density based hierarchical clustering (4h)
Pattern mining and Association Rules: basic concepts and a-priori algorithm (2h)
Practicals: hands-on on simple case studies using python libraries (2h)
3) Supervised Learning: methods and practicals (22 hours)
Classification: introduction, performance evaluation. A first simple classifier: Decision tree (4h)
Practicals: hands-on on simple case studies using python libraries (2h)
Overview of advanced methods: Random Forest, Support Vector Machine (2h)
Introduction to Neural Networks and Project description and project assignment (2h)
Practicals: hands-on on advanced classification methods and Neural Networks with pytorch (2h)
Convolutional NN, theory and practice with Pytorch (2h)
Recurrent Neural Networks (2h)
Generative networks (4h)
Transformers (2h)
4) Design principles and Trustworthy issues on AI systems based (4 hours)
Design guidelines, bias discovery and explainability (2h)
Project seminar and discussion (2h)

Syllabus: Teaching will primarily take the form of lectures in Italian, using an interactive approach to help students understand the fundamental concepts of computational linguistics and machine translation. The core texts by Jezek & Sprugnoli (2023) and Jurafsky, Dan, and Martin, James H. (2026) will be analysed and discussed, supplemented by recent research articles provided during the course.
Exercises involving the critical analysis of specialised texts and case studies will encourage independent thinking and the development of original ideas.
Practical lab sessions will also provide hands-on experience with tools for automatic text analysis and the evaluation of neural machine translation.

Syllabus: SYLLABUS/CONTENT
The course aims to:Introduce the concepts of minimization algorithms for a scalar functional (the loss function).
Provide the necessary tools for practical course execution, such as Python, Tensorflow, and Pytorch.
Cover dense neural networks and examples of their applications in physics.
Explore convolutional neural networks and examples of their applications in physics.
Discuss recurrent neural networks and examples of their applications in physics.
Investigate graph neural networks: inductive bias and examples of their applications in physics.
Examine attention mechanisms: transformers and examples of their applications in physics.
Study generative neural networks and examples of their applications in physics.
Provide an overview of differentiable programming.
The course encompasses these topics to provide students with a comprehensive understanding of machine learning algorithms in the context of physics applications.

Syllabus: PDAI1 introduces students to the fundamental principles of structured programming, with basic applications to data processing. It starts from basic notions of programming (variables, data types, collections, control & repetition structures, functions & modules), and progresses to basic data processing functionalities (loading, manipulation, and visualization of CSV data).

Syllabus: Real-world network characterization:
– Big graph data and social, information, biological and technological networks
– The architecture of complexity and how real networks differ from random networks: node degree and long tails, social distance and small worlds, clustering, and triadic closure.
– Comparing real networks and random graphs. The main models of network science: small world and preferential attachment.
– Assortativity and homophilic behaviors.
– Strong and weak ties, community structure, and long-range bridges.
– Network beyond pairwise interactions: high-order network modeling.Applications:
– Robustness of networks to failures and attacks.
– Dynamic Network modeling.
– Dynamic Community Discovery.
– Link Prediction
– Cascades and spreading.
– Network models for opinion dynamics and epidemics.

Syllabus: � Theory of Visualization:
� Perception and vision processes
� Visual Analytics Process
� Strategies and best practices for effective data visualization
� Taxonomy of Data Visualizations:
� Hierarchical and structured data
� Relational and graph-based data (social networks)
� Temporal and spatial data
� Spatio-temporal data
� Unstructured information (text)
� Technologies & Tools:
� Overview of development environments and graphic libraries for the web (e.g., D3.js, Vue.js, Node.js, Crossfilter, ColorBrewer, Plotly, Vega).
� Case Studies & Project Work:
� Analysis of real-world scenarios and interactive visual web applications (traditionally based on challenges like the VAST Challenge).

Syllabus: � Lecture 1: Introduction, What does it mean to know something?
� Lecture 2: Justification and epistemic norms.
� Lecture 3: The Structure of Knowledge
� Lecture 4: Epistemic Risk, Fallibility, and Standards of Knowledge.
� Lecture 5: The social dimensions of knowing.

Syllabus: 1 – What is cognitive science?
2 – Computationalism and its discontents
3 – Reduction and emergence
4 – Explanation(s) in cognitive science
5 – Theory, models, and effects

Syllabus: 1. Introduction, discussion and choice of specific topics. What is science?
2. Howmany sciences? The method(s) of science. Exact and inexact sciences.
3. Theories, models, data. Experiments and observations.
4. Inferences in science. Falsification, confirmation, disconfirmation.
5. Bayesian rationality and scientific reasoning.
6. Science, pseudoscience, junk science.
7. History of science and scientific progress. The aim(s) of science.
8. Trust and objectivity in science. The role of experts.
9. Social and human sciences.
10. Science, truth, and reality

Syllabus: – Lecture 1 : Introduction � Science, Values, and Objectivity.
– Lecture 2: The Gap Argument
– Lecture 3: Inductive Risk
– Lecture 4: Normative and Thick Concepts
– Lecture 5: Managing values in science

Syllabus: https://www.dropbox.com/scl/fi/6peb6sq8emcyasiacwcg2/Syllabus_causal_inference_macro.pdf?rlkey=zri2kyb4h05ed5ovq919nfsbo&dl=0

Syllabus: Introduction to programming in Python; Introduction to popular Python libraries for ABMs; agent-based computational economics (why? what? how?); applications of agent-based models to macroeconomics.

Syllabus: The course content covers the following topics:

1. Phase field modeling and simulation of damage occurring in human vertebra after screws fixation procedure. Fractures inside the human vertebra can be challenging to analyze and treat due to their complex shapes and locations. The phase field modeling can help simulate the fracture process and shedding light on the underlying mechanisms involved.

2. Exploring novel 3-D in vivo bioprinting technology inside the gastrointestinal system: perspectives and new frontiers. For most of the digestive system diseases, endoscopy is the most basic clinical and effective method. Now scientists aim to treat such problems by exploring a new frontier in 3-D printing: depositing living cells directly inside the human body. The development of a computational model for gastrointestinal mechanics leads to patient-specific in-silico tools unveiling the principles governing food digestion and disease mechanisms at different scales. The course will provide a numerical framework for the treatment of the gastro-intestinal system. In particular the Holzapfel-Ogden model coupled with electrophysiology will be discussed and analyzed.

3. Nanoparticle technology for cancer therapy is a growing area of cancer nanomedicine because of the potential for localized and targeted destruction of cancer cells. Localized effects are dependent on many factors, including nanoparticle size, shape, stiffness and surface properties. Computational modeling is an important tool for investigating and optimizing these parameters. The present course will provide coupled Lattice-Boltzmann Immersed boundary methods for the simulation of transport of deformable cells and nano-particles in three dimensional microcapillary flows and how to model the receptor-ligand mediated adhesion dynamics of deformable nanoparticles in whole blood capillary flow.

Syllabus: The course will cover an overview to the following enabling digital technologies:
– digital twins and virtual testing
– big data analytics
– artificial intelligence and automation for quality control
– additive manufacturing – augmented reality
– cybersecurity
– high performance computingPractical applications are presented for each digital technology with also a holistic perspective, along with its impact on a lean factory based on key performance indicators of industrial relevance.

Syllabus: The course will cover the following fundamentals of academic entrepreneurship:
– the players of research and innovation and their roles (1h, Marco Paggi)
– methods to foster innovation (R&D contracts, public-private joint labs, patents, university spin-offs, public-private joint Ph.D. programs) (1h, Marco Paggi)
– university spin-offs (1h, Marco Paggi)
– elevator pitches (2h, Marco Paggi)
– triple helix model of innovation (1h, Francesco Biancalani)
– business plans (1h, Francesco Biancalani)
– testimonials presenting case studies of academic entrepreneurship (3h, Francesco Biancalani)

Syllabus: – Overview of funding schemes to support research mobility
– Fundamentals of Intellectual Property Rights (patents, copyrights, etc.)

Syllabus: 1. Reinforcement Learning. Basic concepts and intuitions. The problem of temporal credit assignment. Early models: The box model and its neural implementation. A taxonomy of existing RL methods.
2. State value estimation methods: Temporal difference (TD) methods. Monte Carlo, n-step TD, and TD-lambda estimation methods.
3. Control methods: Q-learning, SARSA algorithms, Q-lambda, and SARSA-lambda.
4. Implementation issues: Implementing the Q-learning algorithm in C. Structuring the program. Data structures and main functions. Examples of applications: The grid world, Inverted pendulum, Amoeba, Shooting targets, Visual tracking, Lunar lander, Crawling robot, Self-driving car.
5. Deep Reinforcement Learning. Deep Q-learning models: DQL, Double-DQL, Dueling-DQL, Rainbow DQL, Recurrent DQL.
6. Policy gradient methods. The REINFORCE algorithm. Actor-critic methods: DDPG, TD3, A3D, SAC, and PPO.
7. Stochastic RL: A statistical approach to reinforcement learning for stochastic environments.
8. Curiosity-based learning: Variational Information Maximizing Exploration learning. Surprise-based Intrinsic Motivation. Intrinsic Curiosity Module. Random Network Distillation.
9. Imitation learning: Behavioral cloning. Iterative policy learning. Inverse Reinforcement Learning (IRL): Feature expectation matching, Maximum Entropy IRL, Guided Cost Learning. Generative Adversarial Imitation Learning. Imitation by diffusion.
10. Model-based RL: How to learn a model. The Dyna approach. Dyna extensions: Deep Dyna-Q, D3Q, Dyna-PINN, MBPO. Tree-search methods: Minimax and Monte Carlo Tree Search. AlphaGo.

Syllabus: 1. Introduction to neural computing. From classical AI to neural networks. Main network structures and learning paradigms. Hopfield networks and associative memories.
2. Unsupervised learning and clustering algorithms. Dimensionality reduction: Principal component analysis and Self-organizing maps. Hard clustering algorithms. Soft clustering algorithms.
3. Supervised learning. From the Perceptron to backpropagation. Applications of multi-layer networks. Performance evaluation metrics. Managing overfitting and class imbalance. Radial Basis Function networks. Physics-Informed Neural Networks.
4. Deep networks. Problems in training deep networks: overfitting and vanishing gradient. Regularization and dropout methods. Autoencoders and convolutional networks.
5. Networks for classification. Examples of CNNs and their key novelties: LeNet-5, AlexNet, VGG-Net, GoogLeNet, ResNet, Inception family, SqueezeNet, DenseNet, SENet, NASNet.
6. Networks for object detection. Two-stage detectors (the R-CNN family). Single-stage detectors (YOLO and SSD). Performance metrics. Pyramid networks (FPN, RetinaNet, the YOLO family).
7. Networks for image segmentation. Semantic, instance, and panoptic segmentation tasks. Fully Convolutional Network, U-Net, SegNet, ICNet, Mask R-CNN, YOLO-ACT.
8. Recurrent neural networks. Backpropagation through time. Language modeling. Word embedding. Machine translation. The attention mechanism.
9. Transformers. Encoder-Decoder architectures. Positional encoding. Self-attention and multi-head attention mechanisms. Masked attention.
10. Semi-supervised learning. K-nearest neighbors. Self-training algorithms. Co-training. One-shot learning. Siamese networks. Zero-shot learning. Contrastive learning.
11. Neural networks for object tracking. Generic object tracking. Tracking by detection: SORT, DeepSORT, FairMOT, BYTEtrack.
12. Generative networks. Generative autoencoders, Generative adversarial networks (GANs), GANs for style transfer. Variational Autoencoders, Diffusion models.

Syllabus: – Introduction to digital twins and their use in technology (1h, Marco Paggi)
– Data-driven models based on artificial neural networks: neurons, activation functions, cost functions, back-propagation algorithm for a simple artificial neural network (1h, Marco Paggi)
– Artificial neural networks for linear and nonlinear classification problems (1h, Marco Paggi).
– Time series networks (1h, Marco Paggi)
– Convolutional neural networks (1h, Marco Paggi)
– From Computer Aided Design (CAD) or real geometrical data (from images, laser scanner, etc.) to Computer Aided Engineering (CAE): how to integrate realistic object geometries in simulation tools (3h, Maria Rosaria Marulli)
– Digital twins for cultural heritage (2h, Maria Rosaria Marulli)
– User element routines for model-based simulations of surface problems (5h, Maria Rosaria Marulli)
– Interfaces between different CAE software (1h, Andrea Mola)
– Integrating functions from software libraries in CAE simulation tools (2h, Andrea Mola)
– Integrating data driven information in numerical models (2h, Andrea Mola)