Lecture hours (starting October 1, 2026) Thu 16:30-18:00, room Oe Fri 16:30-18:00, room Me
In classroom teaching This year the course unit is delivered in classroom! For all the students waiting for a visa permit: contact me via email for the zoom link.
Le2 - Fri Oct 2, 2026 Networks as graphs; Directed and undirected networks; Adjacency matrix; Nodes degrees and the degree distribution; Sparsity; Bipartite graphs; Projections on semantic graphs; Signed graphs.
Le3 - Thu Oct 8, 2026 Layered networs; Paths, cycles, diameter, and small world networks; Connectedness and the condensation graph.
Le3 - Thu Oct 8, 2026 Degree centrality; Power law versus Poisson distribution; Estimation methods for the degree exponent: ML approach; Explaining the power law; Erdos-Renyi random model: binomial and Poisson descriptions of the model; Small world; Power law versus Poisson distribution; Growth and preferential attachment; the Barabasi-Albert model.
Le4 - Fri Oct 9, 2026 Mathematics of the Barabasi-Albert model; Attractiveness and the Bianconi-Barabasi model; Analysis of the Bianconi-Barabasi model; Examples with equal and uniform fitness; Measuring fitness: the www; Other ideas for modelling the power-law; Properties of the power-law: largest hub, moments and scale-free networks.
Le5 - Thu Oct 15, 2026 Wrap-up on scale-free networks, distances; PageRank: random walk, dead ends and spider traps, teleportation; PageRank with restart; PageRank versus degree centrality; convergence properties of PageRank; PageRank matrix structure via the condensation graph; Action of the teleport vector: one leaf-only, one eigenvalue set to 1 and the rest to (at most) c; Interpretation.
Le6 - Fri Oct 16, 2026 Tuning the teleport vector: Local PageRank; Examples of application; Approximate PageRank and the push operation: proof of the precision guarantee; Application to link prediction; Topic specific PageRank; Signed PageRank; Preventing spamming; Row-normalized PageRank.
Le7 - Thu Oct 22, 2026 HITs centrality; Eigenvector and Katz centralities and their relation to PageRank and Degree centrality; Closeness and harmonic centrality; Betweenness centrality as a measure of brokerage; The clustering coefficient; Wrap-up on centrality measures; Homophily and related concepts: Selective exposure, Polarization, Ego Chamber.
Le8 - Fri Oct 23, 2026 Filter bubble; Degree homophily or assortativity; the correlation matrix; Nearest neighbour degree function and its exponent; Structural disassortativity; Structural and natural cutoffs; Random rewiring; Network robustness; Robustness of scale-free networks; Attack tolerance; Robustness optimization; An application example; The layout problem; Aesthetic criteria; Spring-embedder algorithm; Repulsive and attractive forces; Fruchterman & Reingold; Force atlas 2; Gravity; Approximate repulsion.
16:30 Intro 16:35 Romanticization of true crime: Dahmer versus Blackbird 16:45 Trends of overconsumption and ProjectPan 16:55 Fast versus vintage fashion 17:05 Social media bots at the time of the french protest 17:15 Online football supporters: club versus national 17:25 Victoria secrets and fashion stereotipes 17:35 Discussion and group forming 18:00 End
Le9 - Thu Oct 29, 2026 Community detection problem: Granovetter's view and the role of weak ties; The core-periphery model and overlapping communities; Modularity: definition and matrix formalization.
Le14 - Fri Nov 13, 2026 Modularity: The Louvain algorithm for modularity optimization and its main characteristics; Solving greediness by consensus clustering; Generalizing modularity: the directed and signed network case; Modularity with overlapping communities.
Le16 - Fri Nov 20, 2026 Modularity with overlap: extending the Louvain approach; Modularity in the two-communities case: spectral approach; The normalised cut criterion; Suboptimal solution to normalised cut: algebraic connectivity and Fiedler's eigenvector; Conductance and the network community profile.
Le17 - Thu Nov 26, 2026 The InfoMap criterion: the node view, the community view, inside communities; Compact result; Normalized mutual information; Wrap-up on community assignment measures; The BigClam: rationale of a model-based approach, algorithm and performance.