Complex Systems Research Exchange (CREx) is an online seminar series aimed at building an international research community in the interdisciplinary field of complex systems, very broadly defined. Our focus is on creating opportunities to hear from early to mid-career researchers about their research, while also serving as a platform for networking and collaboration. We are open to a wide range of research areas including network science, nonlinear dynamics, dynamical systems, computational social science, science of science, machine learning and AI, data science, urban science and human mobility, mathematical biology and ecology, neuroscience, and more.
Please subscribe to the mailing list to receive announcements of upcoming seminars. The Zoom link for each seminar will be sent to the mailing list shortly before the seminar.
Please reach out to organizers if you are interested in giving a presentation yourself or know someone who would be interested. This seminar series is run on a volunteer basis and is currently organized without funding.
Organizers: Jeehye Choi, Takayuki Hiraoka, Inho Hong, Hyewon Kim, Kazuki Nakajima, Ayumi Ozawa, Taekho You
Juhee Lee (Umeå University)
Theoretical approach to stochastic group resetting dynamics
We present a general theoretical framework for group resetting dynamics in multi-particle systems, where the resetting is governed by the collective behavior of the group, unlike traditional models focused on single-particle dynamics. By combining renewal theory with extreme value statistics, we derive a Fokker-Planck equation describing the motion of the group’s center of mass as an effective particle. This allows us to analytically evaluate key quantities such as the stationary mean position, variance, and the squared coefficient of variation (SCV), and to quantify how the group size and resetting rate influence the probability of avoiding a dangerous region. Our framework offers novel insights for designing optimal group-level avoidance strategies, with potential applications in multi-agent optimization, suppression of antibiotic resistance in bacterial populations, and risk mitigation in financial systems.
More talks are coming up! Stay tuned.
Yueran Duan (China University of Geosciences)
Postdoc publications and citations link to academic retention and faculty success
Postdoctoral training is a career stage often described as a demanding and anxiety-laden time when many promising PhDs see their academic dreams slip away due to circumstances beyond their control. We use a unique data set of academic publishing and careers to chart the more or less successful postdoctoral paths. We build a measure of academic success on the citation patterns two to five years into a faculty career. Then, we monitor how students' postdoc positions—in terms of relocation, change of topic, and early well-cited papers—relate to their early-career success. One key finding is that the postdoc period seems more important than the doctoral training to achieve this form of success. This is especially interesting in light of the many studies of academic faculty hiring that link Ph.D. granting institutions and hires, omitting the postdoc stage. Another group of findings can be summarized as a Goldilocks principle: it seems beneficial to change one's direction, but not too much.
Read Yueran's work here.
Andrei A. Klishin (University of Hawaiʻi at Mānoa)
Data-driven models: statistical mechanics, error bars, and failures
The quest for data-driven models of complex systems is usually focused on finding the "best" model by adopting the tools and paradigms of machine learning. Here I switch the emphasis and investigate how training data induces an uncertainty in both model form and model parameters, and how we can navigate that uncertainty with tools of statistical mechanics across two problems. The first problem is recovering the governing equations from observed trajectories of dynamical systems. We recast the popular SINDy method in Bayesian inference terms, interrogate the fit-sparsity trade-off, and visualize the anatomy of equation recovery failure under noise. The second problem is sparse sensing, or reconstructing a full system state from observing just a few well-placed sensors. While conventional methods return the "best" sensor configuration, we show how the configuration naturally arises from the effective interactions between the sensors. In addition, we interrogate and mitigate the "double descent" phenomenon in which the sensor-based reconstruction gets highly unstable at particular sensor numbers.
Based on arXiv:2307.11838 and arXiv:2403.01723
Yuya Sasaki (Osaka University)
Fairness in graph data analysis
Fairness in data science is a critical issue in various applications and has become one of the hot topics in research. In this talk, I will introduce the topic of fairness in graph data analysis. I will explore the types of biases present in graph data, review existing research, and conclude with a discussion of open problems in this area.
Thomas G. de Jong (Kanazawa University)
The Omnipresent RC: Harnessing all Kuramoto-like systems as computational resource
Nature is pervaded with oscillatory behavior. In networks of coupled oscillators patterns can arise when the system synchronizes to an external input. Hence, these networks provide processing and memory of input. We design a Reservoir Computing (RC) framework [1] for harnessing large oscillator networks as a computational resource, referred to as Omnipresent RC [2]. The Omnipresent RC consists of a network of Kuramoto-like oscillators which are forced by an input.
Our results are three-fold. Firstly, the Omnipresent RC inherits performance properties of Kuramoto-like systems, specifically parameters for synchronization are abundant and since the trained network learns via synchronization the Omnipresent RC is generically successful. Secondly, its learning, which describes a type of collective intelligence, can be explained using Kuramoto model's order parameter. Thirdly, when the network is given by the Kuramoto model its continuum limit can be studied under the so-called Ott-Antonsen ansatz [3]. In this continuum limit it turns out that the governing equations that can memorize complicated time series are fairly low dimensional. Additionally, analysis of the parameter space reduces to classical bifurcation theory. In summary, this work provides the foundation for utilizing nature's oscillator networks as a new class of information processing systems.
[1] K. Nakajima (2020), Physical reservoir computing—an introductory perspective. Japanese Journal of Applied Physics 59(6)
[2] T. de Jong, H. Notsu, K. Nakajima (2025), Harnessing omnipresent oscillator networks as computational resource. arXiv:2502.04818
[3] E. Ott, T.M. Antonsen (2008) Low dimensional behavior of large systems of globally coupled oscillators. Chaos 18(3) (Preprint)
Ayumi Ozawa (JAMSTEC)
Disappearance of the collective oscillation in a population of coupled oscillators with turnover
Interacting oscillatory systems can exhibit collective oscillation. This phenomenon, called synchronization, is observed in various systems, such as biological, social, and artificial systems. Although some of these systems experience turnover, i.e., the replacement of the old elements with new ones, most theoretical studies on coupled oscillators have focused on the case where the constitutive oscillators are fixed. In this talk, I will introduce a simple model of coupled oscillators with turnover, where the effect of turnover is described using stochastic resetting. The analysis of the model indicates that as the turnover rate increases, the collective oscillation disappears via two transitions: desynchronization and stochastic oscillation quenching. The talk is based on [1].
[1] Ozawa, A. and Kori, H. (2024) ‘Two Distinct Transitions in a Population of Coupled Oscillators with Turnover: Desynchronization and Stochastic Oscillation Quenching’, Physical Review Letters, 133(4), 047201. (Preprint)
Bukyoung Jhun (IT:U)
Inference of the underlying network structure from observed dynamics
The dynamical states of agents within a system often arise from pairwise interactions embedded in an underlying network structure. However, in many cases—such as epidemic spread, neuronal activation, and social media dynamics—the network of interactions cannot be directly observed. As a result, inferring network structure from observed dynamics has become a central focus in network science. In this talk, we introduce a Bayesian network reconstruction approach based on the stochastic block model. By using the stochastic block model as the underlying model for real-world networks, we significantly narrow the search space of the reconstruction algorithm, enhancing both efficiency and accuracy.
Kazuki Nakajima (Tokyo Metropolitan University)
Exploration of Higher-Order Community Structure in Grant Collaboration Networks Using Stochastic Block Models
Grant collaboration among researchers or institutions has become increasingly common and essential for effective scientific research. It is not uncommon for three or more researchers or institutions to collaborate on a single research grant, and a set of such grant collaborations can be represented as a hypergraph. In this talk, I introduce our ongoing research on the inference and visualization of higher-order community structure. We use a framework that combines a stochastic block model with a dimensionality reduction method to explore community structure in grant collaboration hypergraphs.
Seungwoong Ha (Santa Fe Institute)
Dynamics of Collective Minds: A Computational Model of News-Comment Dynamics
In an online community, many of the community users collectively consist of a unique set of interests and beliefs, which can be expressed as a 'collective mind' of the community. We are interested in what the collective mind of community users is aiming for, especially in terms of diverse topics, and what would be the dynamics of the frequency of the comments on certain topics (along with the similarity between those topics). After gathering data from nearly 10 years of news and comment data from 5 different online news communities, we aim to build a simple yet comprehensive computational model for the online news community where news articles and comments are posted daily. In this model, the outside world that generates 'news', the community-specific filter then determines which news will get posted or not on their community, the 'collective mind' of the community reacts to the filtered news and generates comments, and finally, these comments alter the collective mind itself, repetitively. With this model, our goal is to explore the variety of interventions that may affect the community mind, such as comment filtering, strong editorial policies, and trolls/bots regulations.
Sakurako Tanida (University of Tokyo)
Elevator Traffic Jam: Why Do They Arrive Together?
Hierarchical order structures are fundamental aspects of life and fascinating topics in non-equilibrium physics. Human groups are more complex compared to other species, and human movement also involves hierarchical structures, such as in transportation systems. Man-made systems are often easier to understand than natural systems, making them suitable for investigating hierarchical order formation. On-demand transportation systems, in particular, can exhibit emergent order from noise, leading to intriguing non-trivial phenomena.
In this context, we focus on the simultaneous arrival of elevators in crowded buildings. By conducting simulations and developing mathematical models, we characterize the movements of elevators. Our research investigates how order can emerge from the noise inherent in arrival times, shedding light on this self-organizing mechanism.
Additionally, I will introduce our ongoing research in the presentation. This research explores the dynamics of not just one or two elevators, but multiple elevators and the inclusion of transfer scenarios, integrating higher-order structures into our analysis. I will also present ideas for mitigating synchronization issues, aiming to improve operational efficiency.
Read Sakurako's works here (or here for the preprint) and here (or here for the preprint).
Sadamori Kojaku (Binghamton University)
Machine learns simplicity in complexity
Recent advancements in generative AI have created highly realistic texts and videos that are often indistinguishable from genuine content, demonstrating its capability to learn the complexities of the world. This talk will explore how AIs---specifically, neural networks---learn complex systems. We will discuss how neural networks build a high-dimensional vector space---an internal representation of the world---that reflects the inherent structure of complex systems. We will delve into the mechanics of neural embeddings, using community detection tasks as a practical example. Then, we will focus on human mobility in science, showcasing a robust and implicit connection between embedding distance and human mobility flow, an interpretable, tangible quantity. Finally, we will illustrate how neural embedding can reveal new perspectives on the structure of complex systems by analyzing the citation dynamics in science, law, and patents.
Xiangnan Feng (Complexity Science Hub)
Core Breaking in Minimum Vertex-Cover Problem
The minimum vertex-cover problem belongs to one of Karp’s 21 NP-complete problems and the six basic NP-complete problems, which has a wide range of applications in the related real networks, such as immunization strategies in networks and monitoring of internet traffic. Core difficulty faced by these technologies is commonly and mainly the high computational complexity induced by the real-world applications. In our research, based on the structural features of the leaf-removal cores, a method named core influence is proposed to break the graphs into no-leaf-removal-core ones, which takes advantages of identifying some significant nodes by localized and greedy strategy. Furthermore, solution space organisation of minimum vertex-cover problem is investigated using the famous König-Egérvary (KE) graph and theorem, in which a hierarchical decomposition mechanism named KElayer structure of general graphs is proposed to reveal the complexity of vertex-cover. We hope our research could draw more researchers' attention on the fundamental mathematical problems.
Read Xiangnan's works here (or here for the preprint) and here (or here for the preprint).
Oh-Hyun Kwon (POSTECH)
The Interplay Between Urban Geography and Human Interaction in the Cities
Urban systems present unique challenges for studying human interactions due to their physical constraints. This research bridges the gap between network science and urban geography by leveraging rich human mobility data. We explore how urban layouts, often characterized by centric structures, restrict and influence human movement patterns. By analyzing mobility flows, we uncover the microstructure of mobility deterrence by distance, revealing the intricate relationship between population distribution and human mobility patterns. Furthermore, we investigate how these mobility patterns themselves can reveal underlying urban structures, including the identification of local activity clusters. This approach offers a systematic framework for analyzing diverse urban datasets, providing novel insights into the interconnectedness of urban geography and human interaction.
Read Oh-Hyun's work here.
Seong-Gyu Yang (Korea Institute for Advanced Study)
Understanding biodiversity in large ecosystems with statistical physics approach
Numerous species coexist in nature, forming stable ecosystems despite competing for shared resources. The field of theoretical ecology has developed niche theory to elucidate the formation and sustenance of ecological systems. While the competitive exclusion principle (CEP) states that species with identical niches cannot coexist, observations in phytoplankton communities challenge this principle by showing diverse coexistence despite limited number of resources.
In this study, we introduce intraspecific suppression as a mechanism, extending a competition-based ecological model to comprehensively address coexistence and understand the high biodiversity. Through integration into the generalized MacArthur's consumer-resource model, we demonstrate how intraspecific suppression enhances biodiversity beyond CEP's predicted limit. Analyzing the relative diversity of coexisting consumers and resource kinds at steady state, we employ the cavity method and generating functional analysis to show analytically how this diversity can surpass unity, depending on the strength of intraspecific suppression. Supported by numerical simulations, we reveal that intraspecific suppression restricts the emergence of dominant species, fostering high biodiversity. Moreover, we explore how the effect of intraspecific suppression varies across different environmental conditions. This work offers a comprehensive framework within niche theory, incorporating intraspecific suppression to reconcile CEP's predictions with observed phenomena in ecological systems.
Read Seong-Gyu's work here.
Minsuk Kim (Indiana University, Bloomington)
Shortest-path percolation on complex networks
In various infrastructural networks serving the transport of people or goods, path-based interactions play a significant role in sustaining their functionality. Thus, it is crucial to understand the robustness of such networks upon path-based perturbations. To tackle this problem, we propose a bond-percolation model describing the consumption and eventual exhaustion of resources of networks. In the model, a pair of origin-destination nodes is randomly selected at a time. If the shortest path distance between the selected pair of nodes is within the maximum budget, all edges along one of the randomly chosen shortest paths are removed from the network. As node pairs are selected, the initially connected network progressively fragments into disconnected clusters. We apply this model to Erdős–Rényi networks and fully characterize it by means of finite-size scaling analysis. With a finite maximal budget, the model displays a percolation transition identical to the one of the ordinary bond percolation. With an infinite budget, the transition is more abrupt than the one of the ordinary bond percolation yet smoother than the one that is displayed by the explosive percolation.
Read Minsuk's work here.
Erika Nozawa (Yamagata University)
Complex systems approach to phase inversion phenomena in food emulsions
Coupled map lattice (CML) is a powerful simulation approach that reproduces well complex and diverse patterns and motions in dynamical phenomena with spatial degrees of freedom. Indeed, CML has simulated various phenomena such as full range boiling, convection with turbulence transition, cloud formation predicting ‘guerrilla rainstorms’, and astronomical formation [1, 2] of transient grand-design spirals.
We proposed a CML for simulating phase inversion processes from fresh cream to butter via whipped cream [3]. It is one of complex systems approaches to the pattern formation and self-organization of diverse food textures appearing in the phase inversion processes. The simulations exhibit two different phase inversion processes at high and low whipping temperatures (WTs). The overrun and viscosity changes in these processes are consistent with those in experiments. The two processes give rise to distinctive spatial patterns of overrun and viscosity, and are characterized on the viscosity-overrun plane which is one of the state diagrams, as the viscosity-dominant process at high WT and the overrun-dominant process at low WT, respectively. The butters obtained in the two processes were of low overrun and viscosity and of high overrun and viscosity, respectively in the butter region, and had, so to say, soft & creamy and hard & fluffy texture patterns. In the presentation, we will design a new texture (fluffy & creamy with moderate firmness) by controlling the cooking parameters of the CML procedures based on the above texture difference. We will also discuss our recent results on the size estimation of air bubbles and butter grains, important factors in texture design, if time permits.
[1] E. Nozawa, "Coupled map lattice for the spiral pattern formation in astronomical objects", Physica D 405 (2020) 132377. (Preprint)
[2] E. Nozawa, "Jammed Keplerian gas leads to the formation and disappearance of spiral arms in a coupled map lattice for astronomical objects", Progress of Theoretical and Experimental Physics 2023 (6) (2023) 063A02.
[3] E. Nozawa and T. Deguchi, in preparation.
Yohsuke Murase (RIKEN)
Indirect reciprocity beyond dichotomic reputation updates
Cooperation is a crucial aspect of social life, yet understanding the nature of cooperation and how it can be promoted is an ongoing challenge. One mechanism for cooperation is indirect reciprocity, with which individuals cooperate to maintain a good reputation. In models of indirect reciprocity, the interplay between an individual's actions and the resulting reputation is governed by a community's social norm. While many theoretical studies have been conducted to find norms that achieve stable cooperation, most previous studies conventionally assumed that reputations are binary values, either 'good' or 'bad.' Whereas this assumption has been widely adopted as a common practice for simplicity, such a simple dichotomy is not always realistic, and it is unclear whether the conclusions obtained for the binary-reputation models are valid for cases with more general and nuanced reputations.
This talk presents two studies addressing these limitations in understanding social norms and reputation systems. The first study [1] explores social norms with ternary reputations ('good', 'neutral', 'bad') using supercomputing, comprehensively analyzing cooperative norms beyond the well-known "leading eight" norms. The second study [2] extends the model by considering reputation updates for passive receivers and introducing stochastic elements. For this extended model, we theoretically obtained the necessary and sufficient conditions for cooperative Nash equilibria. Both studies reveal the common rules for successful norms and uncover norms with intriguing and counter-intuitive behaviors. These findings offer insights for designing effective norms in more realistic and nuanced reputation systems.
[1] Yohsuke Murase, Minjae Kim, Seung Ki Baek "Social norms in indirect reciprocity with ternary reputations" Scientific Reports, 12, 455 (2022)
[2] Yohsuke Murase, Christian Hilbe "Indirect reciprocity with stochastic and dual reputation updates" PLOS Computational Biology 19(7): e1011271 (2023)
Jisung Yoon (Northwestern University)
Unsupervised embedding of trajectories captures the latent structure of scientific migration
Human migration and mobility drives major societal phenomena including epidemics, economies, innovation, and the diffusion of ideas. Although human mobility and migration have been heavily constrained by geographic distance throughout history, advances and globalization are making other factors such as language and culture increasingly more important. Advances in neural embedding models, originally designed for natural language, provide an opportunity to tame this complexity and open new avenues for the study of migration. Here, we demonstrate the ability of the model word2vec to encode nuanced relationships between discrete locations from migration trajectories, producing an accurate, dense, continuous, and meaningful vector-space representation. The resulting representation provides a functional distance between locations, as well as a "digital double'' that can be distributed, re-used, and itself interrogated to understand the many dimensions of migration. We show that the unique power of word2vec to encode migration patterns stems from its mathematical equivalence with the gravity model of mobility. Focusing on the case of scientific migration, we apply word2vec to a database of three million migration trajectories of scientists derived from the affiliations listed on their publication records. Using techniques that leverage its semantic structure, we demonstrate that embeddings can learn the rich structure that underpins scientific migration, such as cultural, linguistic, and prestige relationships at multiple levels of granularity. Our results provide a theoretical foundation and methodological framework for using neural embeddings to represent and understand migration both within and beyond science.
Read Jisung's work here.
Dahae Roh (Engineers and Scientists for Change)
Growing hypergraphs with preferential linking
A family of models of growing hypergraphs with preferential rules of new linking is introduced and studied. The model hypergraphs evolve via the hyperedge-based growth as well as the node-based one, thus generalizing the preferential attachment models of scale-free networks. We obtain the degree distribution and hyperedge size distribution for various combinations of node- and hyperedge-based growth modes. We find that the introduction of hyperedge-based growth can give rise to power law degree distribution $P(k) \sim k^{−\gamma}$ even without node-wise preferential attachments. The hyperedge size distribution $P(s)$ can take diverse functional forms, ranging from exponential to power law to a nonstationary one, depending on the specific hyperedge-based growth rule. Numerical simulations support the mean-field theoretical analytical predictions.
Read Dahae's work here (or here for the preprint).