Nino Antulov-Fantulin is a senior researcher at the ETH Zurich (COSS group) and a head of research at Aisot Technologies AG.
At ETH Zurich, he works as a lecturer on the following courses: Data Science in Techno-Socio-Economic Systems; Complex Social Systems: Modeling Agents, Learning, and Games; Complexity and Global System Science; Machine Learning and Modelling for Social Networks.
His main interests are at the intersection of complexity science and (financial) data science. In particular predictive analytics for finance, dynamical processes on networks, machine learning, social network analysis, network dismantling, Monte-Carlo algorithms. His interdisciplinary research contributions include: Nature Commun. 13, 333 (2022), Phys. Rev. Research 2, 033121 (2020), Proc. of the national academy of sciences 116.14 (2019), Pattern Recognition 82 (2018): 40-55., Phys. review letters 114.24 (2015), and different conferences: ICDM 2018, ICML 2019, ICLR 2020.
Beside ETH Zurich, he worked at the Rudjer Boskovic Institute and the Faculty of Electrical Engineering and Computing in Croatia and as a visiting scientist at the Robert Koch Institute (Berlin) & Courant Institute of Mathematical Sciences (New York). He worked on several EU projects: SoBigData - "Social Mining & Big Data Ecosystem", Multiplex−“Foundational Research on MULTI-level comPLEX networks and systems”, FOC−“Forecasting Financial Crisis” and e-Lico− “An e-Laboratory for Interdisciplinary Collaborative Research in Data Mining and Data-Intensive Science”. He also works as Supervisor & Panel member of PhD Program in Data Science, Scuola Normale Superiore, Pisa.
He acts as reviewer for IEEE, ACM, Nature Communications, Nature Scientific Reports, Applied Network Science, ICML, ICLR, ECML-PKDD, NeurIPS and is in the program committee of "International Conference on Complex Networks and Their Applications" and "ECML-PKDD Applied Data Science Track".
Latest Projects & News
Numerous factors affect the spread of coronavirus. To better predict the progression of the epidemic, researchers at ETH Zurich and the University of California, Los Angeles have started a “datathon”. Participants from around the globe are called on to develop models based on publicly accessible data.
See link to ETH news
Press release about our work
Researchers at ETH Zurich and the Frankfurt School have developed an artificial neural network that can solve challenging control problems. The self-learning system can be used for the optimization of supply chains and production processes as well as for smart grids or traffic control systems.
See link to ETH news
Science Animation of SIR spreading simulation
The animation is giving a short introduction to the efficient and simple framework that maps the SIR dynamics to an ensemble of weighted networks using a Monte Carlo approach. The same framework is used to describe: (i) Markovian and non-Markovian processes, (ii) continuous and discrete time processes, (iii) finite and infinite-time behavior, (iv) mean-field and exact process representation, (v) MCMC and independent realization sampling and has connections to percolation and disordered network theory.
Statistical embedding for directed graphs
We propose a novel statistical node embedding of directed graphs, which is based on a global minimization of pairwise relative entropy and graph geodesics in a non-linear way. Each node is encoded with a probability density function over a measurable real n-dimensional space. Furthermore, we analyze the connection to the geometrical properties of such embedding and characterize the curvature of the statistical manifolds. Extensive experiments show that our proposed embedding is better preserving the global geodesic information of graphs, as well as outperforming existing embedding models on directed graphs in a variety of evaluation metrics, in an unsupervised setting. See https://arxiv.org/abs/1905.10227
Response to the blog post by Petter Holme
In this short report, we respond to the blog post about the Generalized Network Dismantling. See report.
Neural-Network Control of Dynamical Systems on Graphs
We study the ability of neural networks to steer or control trajectories of dynamical systems on graphs. In particular, we introduce a neural-network control (NNC) framework, which represents dynamical systems by neural ordinary different equations (neural ODEs), and find that NNC can learn control signals that drive networked dynamical systems into desired target states. To identify the influence of different target states on the NNC performance, we study two types of control:(i) microscopic control and (ii) macroscopic control. Microscopic control minimizes the L2 norm between the current and target state and macroscopic control minimizes the corresponding Wasserstein distance. We find that the proposed NNC framework produces low-energy control signals that are highly correlated with those of optimal control. Our results are robust for a wide range of graph structures and (non-) linear dynamical systems.
Unifying continuous, discrete, and hybrid susceptible-infected-recovered processes on networks
Waiting times between two consecutive infection and recovery events in spreading processes are often assumed to be exponentially distributed, which results in Markovian (i.e., memoryless) continuous spreading dynamics. However, this is not taking into account memory (correlation) effects and discrete interactions that have been identified as relevant in social, transportation, and disease dynamics. We introduce a framework to model continuous, discrete, and hybrid forms of (non-)Markovian susceptible-infected-recovered (SIR) stochastic processes on networks. Providing a unifying description of general Markovian and non-Markovian disease outbreaks, we instead show that the mean transmissibility produces the same phase diagrams independent of the underlying interevent-time distributions.
See full paper
The functioning of many socio-technical systems depends on the ability of its subcomponents or nodes to communicate or interact via its connections, but high connectivity may imply problems. By removing or deactivating a specific set of nodes, a network structure can be dismantled into isolated subcomponents, thereby disrupting the (mal)functioning of a system or containing the spread of misinformation or an epidemic. The researchers (Ren, Gleinig, Helbing, Antulov-Fantulin) at the ETH Zurich recently proposed and published insights about Generalized Network Dismantling. See EU SoBigData blog post, ETHZ headline news , ACM TechNews .