The course focuses on advanced algorithms in the field of network analysis. Lectures will be devoted to the analysis and nature of each algorithm in order to assess the suitability of the method in its application. Experiments with algorithms, tools and selected datasets are carried out in exercises and homework.

Network Analysis Methods 2 - full-time study

doc. Mgr. Miloš Kudělka, Ph.D.

Points for tasks and activities

Participation in seminars and continuous activity: 19-36 marks; credit paper: 10-20 points; Implementation of selected algorithm(s) with a simple user interface to set up experiments: 12-24 marks; Analysis of a larger (real-world) network and report with results: 10-20 marks.

Organization of teaching

PDF

Organizational guidelines for the semester, topics, literature, tools.

Data Structures for Network Representation

PDF

Different representations, storage, tasks and their complexity. Large networks and computational issues.

The seminar implements DoK and works with large networks.

Large and dynamic networks

PDF email data

Different types of networks, properties, tasks.

The seminar works with two temporal data sets.

Link prediction

PDF Networks

Link prediction in networks. Different approaches, methods based on local similarity.

Use and compare different methods based on similarity and common neighbor analysis.

Multilayer networkss - basic information

PDF AUCS network

Multilayer networks as a unifying model for different types of networks. Overview, basic measures (centralities).

Experiments with a multilayer network, computing degree-based centralities.

Multilayer networks - measures and projections

PDF

Distance-based measures, random walk applications. Flattening and projection of multilayer networks.

Implementing a random walk in a multilayer network and using it to compute occupation centrality and more.

Multilayer networks - visualization, communities

PDF

Visualization of multilayer networks. Three approaches to community structure detection.

Visualization using one of the recommended libraries. Implementation of community detection using network flattening.