NetSci 2017 will continue the tradition of offering a School suitable for students or faculty interested in learning about specific network science topics. Anyone registered for the conference may attend any of these lectures - no special registration will be required.
The School is organized by Co-chairs Santo Fortunato and M. Ángeles Serrano.
Alex Arenas of Universitat Rovira i Virgili
This lecture will review the basics of network science and its purposes. We will pay attention on both the structural characterization of networks (degree distribution, distances, clustering, correlations, etc) and how structural attributes of real networks, such as the scale-free property or the small-world phenomenon, influence the dynamical behaviors that take place on them. To round off this introductory talk, we will introduce the main current challenges of network science such as time-varying and multilayer networks. Multilayer networks are attracting large interest because they describe complex systems in formed by several networks indicating interaction of different nature. Examples are ubiquitous from infrastructure to transportation and biological networks. We will describe the state of the art for characterizing and modelling the structure of multilayer networks and for studying their robustness properties.
Contagion and spreading processes on networks
Alessandro Vespignani of Northeastern University
This lecture will provide an introduction to the basic theoretical concepts and tools needed for the analysis of dynamical processes taking place on networks. Topics covered will include: navigation, and exploration processes of complex networks; epidemic spreading; social contagion; computational modeling approaches to contagion dynamic; and reaction-diffusion processes on networks.
Maximum-entropy methods for financial and economic networks
Diego Garlaschelli of Leiden University and Oxford University
This lecture focuses on various methods, all based on the Maximum Entropy Principle, that have been recently proven very successful in analysing, modelling and/or inferring real-world economic and financial networks. As a first example I will discuss various methods for reconstructing financial networks from partial information and estimating the associated level of systemic risk, a problem of great importance for monitoring financial stability. As a second example I will discuss novel models of the international trade network that aim at bridging the gap between the traditional gravity model in economics and more recent network models. Then I will move on the to the identification of early-warning signals of upcoming crises in interbank networks. Finally, I will discuss a method, based on Random Matrix Theory, to detect groups of correlated financial entities from empirical correlation matrices. Throughout these applications, I will stress the importance of properly taking the heterogeneity of the system into account. Failing to do so systematically results in a poor performance of the methods.
Controlling complex networks
Raissa D'Souza of University of California Davis
Our understanding of the collective behaviors of complex networks has matured considerably in the last decade, and cutting edge research efforts now often focus on how to control behaviors of complex networks. This has become a prominent issue cutting across disciplines from engineering, to biology, to social systems. Three main threads include structural controllability, control of nonlinear systems, and how to influence and nudge behaviors in social systems. Here we will survey progress in all three of these areas, understanding that the context is essential and that each area has its own strengths and limitations. Structural controllability exploits the deep connection between graph combinatorics and linear algebra, making it possible to answer control related questions relying on network structure only. Yet, it requires that the dynamics evolving on the network is linear. Work on control of non-linear dynamics on networks typically focuses on the phase space portrait of the dynamics and on exploiting non-linear flows to efficiently move the system between different basins of attraction. Yet, it requires detailed knowledge of complex attractors which are typically not fully understood. Finally, work on control of social systems is fraught with difficulties from the fact that agents are typically not rational to the limited knowledge we have of the system. Yet, simple mathematical models of opinion dynamics provide a starting lens.
Learning, Mining, and Networks
Tina Eliassi-Rad of Northeastern University
In this lecture, we will cover some of the most popular supervised and unsupervised learning algorithms on networks. Under supervised learning, we will cover within- and across-network classification of nodes. Under unsupervised learning, we will cover algorithms for finding patterns and anomalies in large-scale networks. We will discuss both generative and discriminative models and will pay special attention to the scalability of the learning algorithms.