Photo credit: Levy Meshulam

Fast and flexible decision-making is critical for networked systems to successfully manage the uncertainty, variability, and dynamic change encountered when operating in the real world. Decision-making is fast if it breaks indecision as quickly as indecision becomes costly. This requires fast divergence away from indecision in addition to fast convergence to a decision. Decision-making is flexible if it adapts to signals important to successful operation, even if they are weak or rare. This requires tunable sensitivity to input for modulating regimes in which the system is ultra-sensitive and in which it is robust. Nonlinearity and feedback in the decision-making process are necessary to meeting these requirements. I will present theory on nonlinear opinion dynamics that enable fast and flexible decision-making among multiple options for multi-agent systems interconnected by communication and belief system networks. I will also introduce the new notion of “excitable decision-making” and the opportunities it affords. The theories provide a principled and systematic means for designing and analyzing decision-making in systems ranging from robot teams to social networks.

This is joint work with Alessio Franci and Anastasia Bizyaeva and based on the papers:
https://ieeexplore.ieee.org/document/9736598
https://arxiv.org/abs/2308.02755
https://epubs.siam.org/doi/10.1137/22M1507826

Naomi Ehrich Leonard is Chair and Edwin S. Wilsey Professor of Mechanical and Aerospace Engineering and associated faculty in Applied and Computational Mathematics and the Program in Biophysics at Princeton University.  She is also affiliated faculty at the Princeton Neuroscience Institute and Founding Editor of the Annual Review of Control, Robotics, and Autonomous Systems. She received her BSE in Mechanical Engineering from Princeton University and her PhD in Electrical Engineering from the University of Maryland.  She is a MacArthur Fellow, member of the American Academy of Arts and Sciences, and winner of the 2023 IEEE Control Systems Award and 2017 IEEE CSS Henrik W. Bode Lecture Prize. She is Fellow of IEEE, IFAC, SIAM, and ASME.  Her current research focuses on dynamics, control, and learning for multi-agent systems on networks with application to multi-robot teams, collective animal behavior, social networks, and other multi-agent systems in technology, nature, and the visual and performing arts. 

Nanoscale systems, such as electronic circuits or biochemical networks, are complex systems where individual components exhibit a noisy behaviour due to nonnegligible microscopic fluctuations (e.g. of thermal origin).

Understanding the rules of interconnection of these individual noisy behaviours into a global dynamics that is physically meaningful is in general  an open challenge where we have to abandon some of the intuitions gathered from our macroscopic world. In other words, there is a need for a new systems & control theory valid at the nanoscopic scale. 

In this talk, we build fragments of this new theory, and show how it applies to a range of systems, from molecular motors to ultra-low-power electronic memories. We show how to leverage recent results of stochastic thermodynamics, so as to extract useful bounds on the minimum energy that needs be dissipated to accomplish tasks, such as driving the system from an initial to a final state (for instance: writing a bit into a memory) reliably and fast.  

Some related articles: 
https://doi.org/10.1103/PhysRevX.11.031064
https://doi.org/10.1063/5.0152883
https://arxiv.org/abs/2110.13050
https://arxiv.org/abs/2208.14506

His research interests include networked dynamical systems, statistical physics, information theory, complex networks and data science. His works have found applications in neurosciences, biochemistry, social sciences, geography, biophysics and electronics.

In this talk we will present some recently computed concentration inequalities for gossip opinion dynamics over random graphs. A network with stubborn and regular agents is generated from a random graph with independent edges. Agents interact pairwise randomly over the generated network. First we derive high-probability bounds for the difference between the expected final opinions of the regular agents in the original gossip dynamics and the expected final opinions in the dynamics over an averaged graph. Then we use these bounds to show how the network topology influences the expected final opinions. We will also present concentration inequalities for time-averaged opinions and transient opinions. With the help of these inequalities, we show almost exact recovery of community detection algorithms based on a single trajectory of gossip opinion dynamics over stochastic block models. The talk is based on joint work with Yu Xing.

On-line navigation devices are widely used by drivers to obtain recommendations on the fastest route to their destination. Despite their wide usage, their impact on the efficiency of the transportation network is not fully understood. In this work, we propose a road traffic model to assess the effects of real-time route recommendations on an origin-destination pair connected by two possible routes. We suppose that a certain percentage of the users has access to a navigation application, which suggests the route with the shortest travel time. After a comprehensive stability analysis determining the asymptotic behavior of the system, we show that routing apps can lead to deterioration of the network performance, corresponding to failure to satisfy user demand. The detrimental effects are stronger when the user demand is high, when the app penetration rate is high, and when the recommendations are affected by delays. Joint work with Tommaso Toso and Alain Kibangou (University of Grenoble, France).

He obtained an engineering degree in applied mathematics (2004) and a PhD in mathematical engineering (2008) from the same university. He has been a visiting researcher at the University of Illinois at Urbana Champaign in 2003-2004, at the National ICT Australia in 2005 and 2006, and at the Massachusetts Institute of Technology in 2006 and 2008. He was a postdoctoral fellow at the Laboratory for Information and Decision Systems of the Massachusetts Institute of Technology 2009 and 2010, holding postdoctoral fellowships of the F.R.S.-FNRS (Fund for Scientific Research) and of Belgian American Education Foundation. He was also resident scholar at the Center for Information and Systems Engineering (Boston University) in 2018-2019, holding a WBI.World excellence fellowship.

Doctor Hendrickx is the recipient of the 2008 EECI award for the best PhD thesis in Europe in the field of Embedded and Networked Control, and of the Alcatel-Lucent-Bell 2009 award for a PhD thesis on original new concepts or application in the domain of information or communication technologies.

The spread of social innovations, technological inventions, viral infections and reforms of corporate governance, are examples of cascading behaviors where the adoption of an action by a portion of individuals makes it more likely for the action to spread to the rest of the population. These population dynamics can be modeled by networked coordination games, and it is of growing interest to study how such games can be controlled effectively over time. In this talk, I show a few analytical tools from evolutionary game theory, Markov decision theory and Q-learning that are handy to establish convergence proofs; I demonstrate that in networked coordination games when decision-making agents follow asynchronous best-response rules, by fixing the behavior of a small portion of “targeted” agents, one may be able to guide the overall population to achieve a pre-specified collective state.

Signed networks model systems where interactions may either be positive or negative, and allow to investigate the emergence of collective, conflictual behaviour. In this talk, I will present a collection of recent results aiming at understanding the relation between the structure and the dynamics of a signed network. The talk will cover the study of diffusive dynamical processes, most notably random walks and consensus dynamics, on signed networks, showing how the balance of the graph impacts their asymptotic and how the dynamical process may help to uncover their mesoscale structure. I will then focus on community detection, discussing how to generalise existing greedy optimisation algorithms in the signed setting, in particular the Louvain method, on the collection of large-scale relational data with signs, and of the randomisation of signed networks for hypothesis testing. I will conclude with possible generalisations in the case when the weight of edges are complex numbers.

When a user shares her data with online platforms, she reveals information about others in her social network. In such a setting, network externalities depress the price of data because once a user’s information is leaked by others, she has less reason to protect her data and privacy. These depressed prices lead to excessive data sharing. We characterize conditions under which shutting down data markets improves welfare. Platform competition does not redress the problem of excessively low data prices and too much data sharing and may further reduce welfare. We propose a scheme based on mediated data sharing that improves efficiency.

This is based on the joint paper with Daron Acemoglu, Ali Makhdoumi, and Asu Ozdaglar.

Reinforcement learning (RL) has had tremendous successes in many artificial intelligence applications. Many of the forefront applications of RL involve multiple agents, e.g., playing chess and Go games, autonomous driving, and robotics. Unfortunately, classical RL framework is inappropriate for multi-agent learning as it assumes an agent’s environment is stationary and does not take into account the adaptive nature of opponent behavior. In this talk, I focus on stochastic games for multi-agent reinforcement learning in dynamic environments and develop independent learning dynamics for stochastic games: each agent is myopic and chooses best-response type actions to other agents’ strategies independently, meaning without any coordination with her opponents. There has been limited progress on developing convergent best-response type independent learning dynamics for stochastic games. I will present our recently proposed independent learning dynamics that guarantee convergence in zero-sum stochastic games. We then focus on the minimal information setting where agents do not observe opponent’s actions, but only observe the payoff they receive at each round. We present payoff-based and independent learning dynamics for such settings and provide finite-time guarantees using a novel coupled Lyapunov drift approach. In the end, I will present a new class of Markov games that models local payoff interactions in multi-agent stochastic games.

Professor Ozdaglar is the recipient of a Microsoft fellowship, the MIT Graduate Student Council Teaching award, the NSF Career award, the 2008 Donald P. Eckman award of the American Automatic Control Council, the 2014 Spira teaching award, and Keithley, Distinguished School of Engineering and Mathworks professorships. She is an IEEE fellow, IFAC fellow, and was selected as an invited speaker at the International Congress of Mathematicians. She received her Ph.D. degree in electrical engineering and computer science from MIT in 2003.

Real world networks — from brain networks to social networks to critical infrastructure networks — are composed of nodes with nonlinear behaviors coupled together via highly non-trivial network structures. Approaches from statistical physics reveal the fundamental implications that complex network structure has on network function and resilience. In contrast, approaches from dynamical systems and control theory reveal the impact that nonlinear nodal dynamics have on emergent behaviors when connected together in simple networks. This talk presents recent work bridging the fields. We show that the interaction between the nodal dynamics and network structure can give rise to novel emergent synchronization behaviors and extend the analysis of cluster synchronization to hypergraphs, capturing higher-order interactions in networks. With respect to cascading failures, we show that adding in oscillatory nodal dynamics to classic models of self-organized-criticality leads to an emergent timescale and the occurrence of self-amplifying dragon king failures that wipe out the system. Finally, we discuss the frontiers of control of complex networks with non-linear nodes, identifying the key challenges and opportunities for bridging control theory, dynamical systems and statistical physics.

Prof. Zampieri has published more than 150 journal and conference papers. He was general chair of the 1st IFAC Workshop on Estimation and Control of Networked Systems 2009, program chair of the 3rd IFAC Workshop on Estimation and Control of Networked Systems 2012 and publication chair of the IFAC World Congress 2011. He was one of the recipients of the 2016 IEEE Transactions on Control of Network Systems Best Paper Award. His research interests include networked control, control of complex systems and distributed control and estimation with applications to the smart grids.

The first case is known as Explosive Synchronization, a phenomenon which has been described in various circumstances where an abrupt onset of synchronization follows an infinitesimally small change in the coupling strength, and hysteresis loops may be observed as in a thermodynamic first-order phase transition.
The second case is the most common one, and corresponds instead to a second-order phase transition
resulting in a sequence of intermediate states in between the completely ordered and disordered phases. Namely, the path to synchrony is characterized by structured states where the network’s nodes split into different functional modules (or clusters) each one evolving in a synchronized manner.

Many studies have pointed out that the underlying graph symmetries are responsible for the way nodes split in functional clusters. In particular, it has been shown that the clusters that are formed during the transition correspond to the equitable partitions of the graph.

In my talk, I will provide a complete elucidation of the transition to synchronization of networked systems.
First, I will describe a (simple, effective, and limited in computational demand) method able to predict the
entire sequence of events that are produced during the transition, and to precisely identify which graph’s node belongs to which synchronized clusters, and to provide a rigorous calculation of the exact coupling strength’s value at which each of such clusters appears.Then, I will demonstrate that such a sequence is in fact a universal graph’s structural property, in that it does not depend on the specific dynamical system operating in each network’s node.
Finally, I will present extensive numerical simulations with both synthetic and real-world networks,which demonstrate how accurate the predictions are.

He was Scientific Attache’ of the Italian Embassy in Israel during the years 2007-2011 and 2014-2018. He is currently Director of Research at the Institute of Complex Systems of the Italian CNR, in Florence.

His major scientific interests are i) pattern formation and competition in extended media, ii) control and synchronization of chaos, and iii) the structure and dynamics of complex networks.

He is Editor in Chief of the Journal “Chaos, Solitons and Fractals” (Elsevier) from 2013, and member of the Academia Europaea since 2016. He was elected member of the Florence City Council from 1995 to 1999.

Boccaletti has published 352 papers in peer-reviewed international Journals, which received more than 34,000 citations (Google Sholar). His h factor is 69 and his i-10 index is 224. With more than 12,000 citations, the monograph ¨Complex Networks: Structure and Dynamics¨, published by Boccaletti in Physics Reports on 2006 converted into the most quoted paper ever appeared in the Annals of that Journal.

Network control systems are the subject of much ongoing research but what about very large-scale sys- tems? From large flocks of minidrones or flying insects to cellular populations, finding a way to close the loop between the microscale, where agents live, and the macroscopic properties that typically need to be controlled is a crucial open challenge for network control. Controlling a system of interest requires being able to sense, compute and actuate. When complex systems are involved, addressing each of these ingredients requires finding new approaches. What and how many agents need to be controlled/sensed and at what scale is a question beyond the current state-of-the-art that requires a new holistic approach to modelling/analysing and controlling large scale network systems. In this talk I will give snapshots into some recent work from my group aimed at addressing the problem of combining concepts from network science, control theory and other disciplines for the solution of this class of problems. I will use some case studies from applications to motivate and illustrate the theoretical derivations giving an overview of the key challenges in the field and opportunities for further research.

The current global financial system forms a highly interconnected network where a default in one of its nodes can propagate to many other nodes, causing a catastrophic avalanche effect. In 2001 Eisenberg and Noe proposed a seminal model allowing to portray the default propagation through a network and evaluate the systemic loss caused by these defaults. In my talk, multi-stage dynamical counterparts of the Eisenberg-Noe model will be considered that have been proposed in our joint works with G. Calafiore and G. Fracastoro. We consider the problem of finding an optimal clearing vector in the dynamic model and prove that this optimal vector is always unique. We also consider a problem of reducing the effect of financial contagion by introducing some targeted interventions that can mitigate the cascaded failures.