Social networks through the ages
Pádraig has a PhD in Applied Mathematics from Coventry University on social networks. After this he did a postdoc in the University of Oxford where he worked on a large European mobile phone dataset. Currently he works in Maynooth University on the Letters of Ireland 1916-23 project.
Online social networks are abundant in our modern lives with media like Facebook and Twitter connecting people all over the world. While studies of modern network structures have allowed us to make observations about human social networks, little work has been done on the structure of pre-internet society.
In this presentation I will talk about applications of network science to three different datasets. The first is ancient and epic narratives, here the structure of social networks from different texts are compared. For example the Sagas of the Icelanders purport to tell the settling of Iceland between the 9th and 11th centuries. The different narratives tend to be loosely connected allowing us to study a large social network comprising multiple texts. This network has many properties similar to modern social networks. Next we look at an early 20th century letters network in Ireland. Although the network is incomplete it still has a large connected component and exhibits similar structural properties to modern communication networks. Finally we contrast the real letters network and quasi-fictional Icelandic network with a network of modern comic book characters. Although this network is entirely fictional we can still make observations about recent changes in society through it.
ORTEC and Network Optimization
ORTEC is one of the world's leaders in optimization software and analytics solutions. We make companies more efficient, more predictable and more effective. We turn complex challenges into easy-to-use solutions.
Since the foundation in 1981, ORTEC has grown into an international company with over 800 ambitious employees worldwide. In the Netherlands about 500 colleagues are active for three business units:
- 'ORTEC Products' develops stand-alone, custom-made and SAP embedded advanced planning and scheduling software. Optimizing everything from fleet routing & dispatch to pallet & space loading, workforce scheduling to warehouse control and delivery forecasting to network planning.
- 'ORTEC Consulting' provides advanced analytics and optimization solutions for companies to survive, innovate and outperform. We offer tailor made and off-the-shelf analytics and optimization models and tools, analytics & consulting services for every level of maturity, as well as experts in the area of data science, business analytics, optimization modeling and software engineering.
- 'ORTEC Living Data' consists of three separate businesses Adscience, imgZine and ORTEC Sports. They each provide ground-breaking software to a particular industry. ORTEC Adscience provides a smarter way to use real-time bidding in performance display advertising. ORTEC imgZine enables publishers, companies and other organizations to publish real-time social magazines based on existing content effortlessly on mobile devices. ORTEC Sports develops software to analyze effectiveness in sport.
ORTEC is one of the world's leaders in optimization software and analytics solutions. We make companies and their networks more efficient, more predictable and more effective. Turning complex challenges into easy-to-use solutions. In this presentation, we will give insight in how we put this to practice. After a generic ORTEC introduction, two customer cases will be presented. We will discuss the added value of mathematical optimization for an aviation company and explain an example of optimizing the supply chain network of a post and parcel company on both tactical and operational level. We are looking forward to present the art of network optimization!
Exploration on Dynamic Networks
Frank Den Hollander is professor in probability theory and statistical mechanics. He is now afliated to the Department of Mathematics at Leiden University, after working in at Eindhoven University of Technology, Radboud University Nijmegen, Utrecht University and Delft University of Technology. He has made breakthrough international discoveries in probability theory, statistical mechanics and network theory. This year he teaches a course on coupling theory, students might also know him for teaching the course Introduction to Probability Theory.
Search algorithms on networks are important tools for the organisation of large data sets. A key example is Google PageRank, which assigns a numerical weight to each element of a hyperlinked set of documents, such as the World Wide Web, with the purpose of measuring its relative importance within the set. The weighting is achieved by exploration. A hyperlink counts as a vote of support. The PageRank of a page is defined recursively, and depends on the number and the weight of all the pages that link to it. A page that is linked to by many pages with a high rank receives a high rank itself.
Real-world networks are modelled as graphs, consisting of a set of vertices and a set of edges connecting pairs of vertices. Complex networks are modelled as random graphs, where the vertices and the edges are chosen randomly, according to an appropriate probability distribution. Search algorithms, in turn, are modelled as random walks, which move along the network by randomly picking an edge incident to the vertex that is currently visited and jumping to the vertex at the
other end. The mixing time of a random walk on a random graph is the time it needs to approach its stationary distribution (also called equilibrium distribution). The characterisation of the mixing time
has been the subject of intensive study. One of the key motivations is that it gives information about the geometry of the graph.
Many real-world networks are dynamic in nature. It is therefore natural to study random walks on dynamic random graphs. In this talk we focus on one particular example, namely, a random graph with prescribed degrees (also called the configuration model). We investigate what happens to
the mixing time of the random walk when at each unit of time a certain fraction of the edges is randomly rewired. We identify three regimes in the limit as the graph becomes large: fast, moderate, slow dynamics. These regimes exhibit surprising behaviour.
Joint work with Luca Avena (Leiden), Hakan Guldas (Leiden) and Remco van der Hofstad (Eindhoven).
Uncovering Offshore Finance using Network Science
Frank is an assistant professor at the Leiden Institute of Advanced Computer Science (LIACS).
He teaches a master course on Social Network Analysis and a bachelor course on Business Intelligence.
His main research interest is in Computational Network Science.
Multinational corporations use highly complex network structures of parents and subsidiaries to organize their operations and ownership.
Offshore Financial Centers (OFCs) facilitate these structures through low taxation and lenient regulation, but are increasingly under scrutiny, for instance for enabling tax avoidance.
Therefore, the identification of OFC jurisdictions has become a politicized and contested issue.
We introduce a novel data-driven approach for identifying OFCs based on the global corporate ownership network, in which over 90 million firms (nodes) are connected through 70 million ownership relations. (Based on: J. Garcia-Bernardo, J. Fichtner, F.W. Takes and E.M. Heemskerk, Uncovering Offshore Financial Centers: Conduits and Sinks in the Global Corporate Ownership Network, Scientific Reports 7, article 6246, 2017. URL (open-access): https://www.nature.com/articles/s41598-017-06322-9)
Uncertainty drivers in the global network of embodied carbon emissions.
João Dies Rodrigues is an Assistant Professor at CML working primarily in Environmental Input-Output Analysis, the field which studies (among other things) how environmental pressures propagate across supply chains. At a theoretical level he develops methods to quantify uncertainty propagation, improve transparency in data processing and promote collaboration and automation in data compilation and validation, using maximum entropy and Bayesian inference among other tools. His applied research focuses on the development of hybrid models (e.g., combining subnational and international multi-regional data, process-based LCA data or mixed unit data) to assess the impact of environmental policies across multiple spatial and sectoral scales. His has both scientific and consulting publications on the topics of energy policy, waste management and carbon footprints/responsibility. He is currently teaching the course of Environmental Input-Output Analysis in the Master Programme of Industrial Ecology.
There is currently wide scientific consensus that anthropogenic greenhouse gas emissions, and CO2 in particular, are major drivers of climate change, with the average global temperature expected to increase by more than 2C by 2100 if significant action is not undertaken. Determining which actions can mitigate carbon emissions and how costs can be fairly allocated requires understanding how the consumption of goods and services drives emissions in production sectors, through the global network of embodied carbon. Substantial effort has already been made to map that network, but most studies provide only point estimates, conveying little information about the uncertainties involved. In this presentation I will describe a study of the uncertainty in the global network of embodied carbon emissions, which uses a set of five global multi-regional input-output models to perform Monte Carlo integration with dependent sampling. After presenting background information I will show our main findings, in terms of prioritizing the elements in the source data that most contribute to global uncertainty, whether general patterns are apparent in the second moments of the source data and the performance of various simplifications. I will conclude the talk with a methodological and epistemological discussion of computational challenges and results.
Network Reconstruction for the Prediction of Spreading Processes
Diego Garlaschelli is Associate Professor at the IMT School of Advanced Studies, Lucca (IT) and at the Lorentz Institute for Theoretical Physics, Leiden University (NL).
He is Associate Fellow of the Saïd Business School, Oxford University (UK). Since 2011, he leads a research group with strongly interdisciplinary interests, including network theory, economic complexity, social dynamics, statistical physics and graph theory. He teaches courses in Complex Networks and Econophysics. He holds a master degree in theoretical physics from the University of Rome III (2001) and a PhD in Physics from the University of Siena (2005). He held postdoctoral positions at the Australian National University in Canberra (Australia), the University of Siena (Italy), the University of Oxford (UK) and the S. Anna School for Advanced Studies in Pisa (Italy). He has given more than 50 invited talks at international conferences, workshops, and scientific schools. He is author of more than 90 publications in peer-reviewed international journals and peer-reviewed book chapters, and of one monograph.
Financial contagion is an epidemic-like phenomenon whereby financial distress can propagate across a network of banks connected by credit relationships, possibly leading to the collapse of the entire system. In order to estimate the systemic risk of financial contagion, the knowledge of the entire interbank network is required. However, due to confidentiality issues, banks only disclose their total exposure towards the aggregate of all other banks, and not their individual exposures towards each bank. A similar problem is encountered in epidemiology. Is it possible to statistically reconstruct the hidden structure of a network in such a way that privacy is protected, but at the same time higher-order properties are correctly predicted? In this talk, I will present a general maximum-entropy approach to the problem of network reconstruction and systemic risk estimation. I will illustrate the power of the method when applied to various economic, social, and biological systems. Then, as a counter-example, I will show how the Dutch interbank network started to depart from its reconstructed counterpart in the three years preceding the 2008 crisis. Over this period, many topological properties of the network showed a gradual transition to the crisis, suggesting their usefulness as early-warning signals. By definition, these signals are undetectable if the network is reconstructed from partial bank-specific information.
Artificial Intelligence, the Machine Learning revolution.
Bert Kappen completed his PhD in theoretical particle physics in 1987
at the Rockefeller University in New York. From 1987 until 1989 he
worked as a scientist at the Philips Research Laboratories in Eindhoven,
Since 1989, he is conducting research on neural networks, Bayesian machine
learning, stochastic control theory and computational neuroscience. His
research has made significant contributions to approximate inference
in machine learning using methods from statistical physics; he has
pioneerd the field of path integral control methods for solving large
non-linear stochastic optimal control problems and their relation to
In 1997, his research was awarded the prestigious national PIONIER
research subsidy from NWO/STW. Since 1997 he is associate professor and
since 2004 full professor at the Radboud University. In 2005 he was Miller
visiting professor at the University of California at Berkeley. Since
2009, he is honorary faculty at UCL's Gatsby Computational Neuroscience
Unit in London.
He co-founded in 1998 the company Smart Research that commercializes
applications of neural networks and machine learning. Smart Research has
developed forensic software for DNA matching used by the Dutch Forensic
institute (MH17 plane crash over Ukraine in 2014), Interpol, the Vietnam
government for analysis of victims of the Vietnam war and the Australian
Police force. He is director of the Dutch Foundation for Neural Networks
(SNN), which coordinates research on neural networks and machine learning
in the Netherlands through the Machine Learning Platform.
Over the last 10 years, artificial intelligence has been revolutionized by deep learning and modern computer architectures. In this talk I will present an overview of some of the activities in this field. I will then focus on control problem: how to compute actions for complex tasks such as robotics and how neural networks can help here. Finally (if time permits) I will comment on the problems of energy use by modern computer systems. Here also we can learn from the brain. I will outline some of our current research plans to build neural circuits at atomic level.
Networks in Neuroscience
Tansu Celikel received his PhD in Systems Neuroscience from the La Scuola Internazionale Superiore di Studi Avanzati (Italy) in 2001. After postdoctoral research at the University of California, San Diego and the Max-Planck Institute for Medical Research, he started his first laboratory at the University of Southern California. Since 2012, Dr. Celikel has been with the Donders Institute for Brain, Cognition and Behavior in the Netherlands, where he is the Founding Chair of the Department of Neurophysiology, and the Founding Director of the Graduate School of Bionics. Dr. Celikel's research is focused on reverse engineering the brain circuits and neural computations that are responsible for sensorimotor control, and forward engineering brain inspired in silico circuits that perform active sensing and mapless navigation. He is a Sloan Fellow, Alexander von Humboldt Fellow, and a Whitehall Investigator.
The brain is made up of an intricate web of networks. From the molecular networks that regulate the cellular organization in the brain to functional networks that orchestrate the dynamics of behavior, common principles of network computation can be observed across all scales of brain formation and function. In this lecture, I will discuss the basic principles of network formation in the brain, introduce molecular, synaptic, cellular and functional brain networks, and provide examples of how robust computation emerges from fail prone network components.
Networks in Pieces
Hans Bodlaender works four days per week as professor Algorithms and Complexity at the Department of Computer Science of Utrecht University, and one day per week at the Department Algorithms and Visualization of the Technical University Eindhoven. His scientific work aims at the design and analysis of algorithms, in particular for networks. Important themes are the use of structural properties of networks to speed up the algorithms, like the existence of a suitable tree structure that represents the network, and the fixed parameter complexity and kernelization: methods to solve problems more efficiently when the input has parameters of which we can expect that they are small. Hans Bodlaender is married, has three children; his hobbies include board games, running and hiking.
In many applications, we have a network (e.g., a road network, a social network, an electrical network) and we want to compute "something" about this network. In this talk, we look how cutting the network in parts can help to speed up such computations. Examples of such applications are the classic method to compute the resistance of an electrical network and recent methods to speed up short routes in a road network. The last example that is discussed is the computation of so-called Shapley values in social networks; these values give an indication of the importance of persons in the network.
What if you don't know it all?
Federica Parisi obtained her master degree in Mathematics from the University of Turin (Italy) with a thesis on Kalman Filter application to atomic clocks prediction and timescale generation. She is a PhD candidate at the IMT School for advanced studies in Lucca (Italy), in the Complex Networks group. Her research interest involve maximum entropy models for network reconstruction and link imputation, network evolution and uncertainty on network observations. She is currently visiting the "Econophysics and network theory" group at Leiden University.
Every kind of network analysis relies on a basic, yet sometimes unrealistic, assumption: the knowledge of the network itself. However, in many real life situations, we might not have full access to the information we need to build a network. Weather it is because of errors, privacy issues or budget constraints we might be able to observe only a subset of the network links, knowing that a part of them is missing. Here we deal we this scenario using maximum entropy techniques to impute those missing links. That is, our goal is to give our best guess on their location given the available information and without any further (possibly biased) assumption.
How well can we reconstruct a bipartite network?
Alexander Becker is a PhD student in the group of Prof. H. Eugene Stanley at Boston University. He uses network science to study foreign exchange markets as well as systemic risk of financial systems. As a visitor in the "econophysics and network theory" group at Leiden University, he is testing network reconstruction methods with empirical data.
As companies take out loans from banks, they form a network in which the failure of an institution can propagate through the system. If we want to understand the risk but don't know the individual connections, we need an effective reconstruction of the network. Maximum entropy methods are a away to impute the links, and using data from Japanese firms and banks, we study how well we can do if we don't know it all.
To be announced
Daniel Rodrigues Valesin is assistant professor at the University of Groningen.
His main research interests are Interacting Particle Systems, Percolation Theory and Random Graphs.
He originally is from Brazil, where he did his Bachelor and Master.
After this he did his PhD in Switzerland and Post-doc at The University of British Columbia in Canada, before he came to Groningen.
To be announced