Registered Data

[00444] Complex Systems: Advances in Theory and Applications

  • Session Time & Room :
    • 00444 (1/2) : 1C (Aug.21, 13:20-15:00) @G304
    • 00444 (2/2) : 1D (Aug.21, 15:30-17:10) @G304
  • Type : Proposal of Minisymposium
  • Abstract : Many social, biological, and technological networks display non-trivial features, with complicated structures patterns of connection. Well-known classes of complex networks are scale-free and small-world networks. The study of complex networks is growing and many new aspects of network structures attract attention in mathematics, physics, electric power systems, biology, climate, computer science, sociology, epidemiology, and others. There is also a wide range of practical issues including Coupled networks and cyber-physical networks; Networked control; Multi-agent systems: Synchronization phenomena; Complex engineering design, including communication networks, power grids, electronic circuits, biomedical systems, software systems; Biological systems, neural networks, disease transmission.
  • Organizer(s) : Maciej Ogorzalek
  • Classification : 05C82, 68Q06, 68M10, 91D10
  • Minisymposium Program :
    • 00444 (1/2) : 1C @G304 [Chair: Guanrong Chen]
      • [01282] Time Series Analysis with Machine Learning
        • Format : Talk at Waseda University
        • Author(s) :
          • Michael Small (University of Western Australia)
          • Braden John Thorne (University of Western Australia)
          • Eugene Tan (University of Western Australia)
          • Débora Corrêa (University of Western Australia)
          • Ayham Zaitouny (University of Western Australia)
          • Thomas Stemler (University of Western Australia)
        • Abstract : Machine learning is widely applied to model dynamical systems and make predictions. Instead of doing this, We will introduce the concept of Reservoir Time Series Analysis - using a particular flavour of machine learning (reservoir computing) to represent the state of a dynamical system and characterise the dynamical evolution of that state. How much can we infer about the changing behaviour of a system from the internal representation of these states within a reservoir machine learning model? A second strategy within machine learning for time series analysis is to use the machine learning model as a proxy for the original dynamics - but how well do such models capture chaotic dynamics? I will show via some short examples that persistent homology can be used as an effective tool to quantify that structure. These methods will be illustrated with applications to machine vibration and pump cavitation in industrial processes.
      • [01306] On constructing directed networks from multivariate time series
        • Format : Talk at Waseda University
        • Author(s) :
          • Tomomichi Nakamura (University of Hyogo)
          • Toshihiro Tanizawa (Toyota)
        • Abstract : We consider a problem of constructing networks for multivariate time series. To construct networks from multivariate time series, we first need to detect relationships among them. However, it is difficult, because the relationships among multivariate time series are diverse. The time series might contain components that have large differences in the amplitude and the time scales of the fluctuations. We consider this problem using the transfer entropy, one of the common techniques for detecting causal relationships and Reduced Auto-Regressive (RAR) model, an information theoretic reduction of auto-regressive model.
      • [05563] Optimal network synchronization and a higher-order topological approach
        • Author(s) :
          • Guanrong (Ron) Chen (City University of Hong Kong )
        • Abstract : In this talk, we will discuss the optimal network synchronization problem. The totally homogenous network approach will be reviewed, and a higher-order topological approach will be introduced, with some preliminary results reported.
      • [01273] Evaluating the Network Robustness: A Convolutional Neural Network Approach
        • Format : Online Talk on Zoom
        • Author(s) :
          • Yang Lou (Osaka University)
          • Junli Li (Sichuan Normal University)
          • Guanrong Chen (City University of Hong Kong)
        • Abstract : Evaluating network robustness by attack simulations is computationally time-consuming. In this talk, a convolutional neural network (CNN)-based robustness estimation is presented, which has three schemes, including the straightforward scheme, the learning feature-assisted scheme, and the pyramid pooling-assisted scheme. Experimental studies demonstrate that: 1) the prediction error is low; 2) the runtime is significantly lower than that of attack simulations; and 3) it provides a good indicator for robustness, better than the classical spectral measures.
    • 00444 (2/2) : 1D @G304 [Chair: Maciej Ogorzalek]
      • [01264] Adaptive Finite-Time Output Consensus for Fractional-Order Complex Networks With Multiple Output Derivative Couplings
        • Format : Talk at Waseda University
        • Author(s) :
          • Chenguang Liu (Beihang University)
          • Qing Gao (Beihang University)
          • Jinhu Lu (Beihang University)
        • Abstract : This paper delves into the adaptive fifinite-time output consensus (FTOC) problem for a multiple output derivative coupled fractional-order complex network (MODCFOCN). Based on the properties of the Gamma function and the fractional derivetive, a FTOC criterion for the MODCFOCN is derived by designing an appropriate adaptive output-feedback controller. Finally, a numerical example is utilized to substantiate the effectiveness of the acquired FTOC results and the devised adaptive output-feedback controller.
      • [01255] Optimizing 3D Complex Networks on chip
        • Format : Talk at Waseda University
        • Author(s) :
          • Maciej Ogorzalek (Jagiellonian University)
          • Katarzyna Grzesiaak-Kopec (Jagiellonian University)
        • Abstract : Current generations of integrated circuits can contain billions of transistors and extremely complicated interconnect network the length of which goes into dozens of kilometers – all these placed in an extremely small physical volume. For the correct operation of the circuit placement of elements and building blocks and design of interconnect has to be done in optimal or quasi-optimal way satisfying also several types of constraints. Mathematical models of complex networks can play a significant role in the design and AI-based methodologies can be used for finding good/quasi-optimal solution in reasonable time.
      • [01268] Machine Learning for Detecting Internet Traffic Anomalies
        • Format : Online Talk on Zoom
        • Author(s) :
          • Ljiljana Trajkovic (Simon Fraser University )
          • Ljiljana Trajkovic (Simon Fraser University)
        • Abstract : Border Gateway Protocol (BGP) enables the Internet data routing. BGP anomalies may affect the Internet connectivity and cause routing disconnections, route flaps, and oscillations. Hence, detection of anomalous BGP routing dynamics is a topic of great interest in cybersecurity. Various anomaly and intrusion detection approaches based on machine learning have been employed to analyze BGP update messages collected from RIPE and Route Views collection sites. Survey of supervised and semi-supervised machine learning algorithms for detecting BGP anomalies and intrusions is presented. Deep learning, broad learning, and gradient boosting decision tree algorithms are evaluated by creating models using collected datasets that contain Internet worms, power outages, and ransomware events.
      • [01280] Model for estimating unconfirmed COVID-19 cases and multiple waves of pandemic progression
        • Format : Talk at Waseda University
        • Author(s) :
          • Choujun Zhan (South China Normal University)
          • Chi K. Tse (City University of Hong Kong)
        • Abstract : The novel coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has unique epidemiological characteristics that include presymptomatic and asymptomatic infections, resulting in a large proportion of infected cases being unconfirmed, including patients with clinical symptoms who have not been identified by screening. These unconfirmed infected individuals move and spread the virus freely, presenting difficult challenges to the control of the pandemic. To reveal the actual pandemic situation in a given region, a simple dynamic susceptible-unconfirmed-confirmed-removed (D-SUCR) model is developed taking into account the influence of unconfirmed cases, the testing capacity, the multiple waves of the pandemic, and the use of nonpharmaceutical interventions. Using this model, the total numbers of infected cases in 51 regions of the USA and 116 countries worldwide are estimated, and the results indicate that only about 40% of the true number of infections have been confirmed. In addition, it is found that if local authorities could enhance their testing capacities and implement a timely strict quarantine strategy after identifying the first infection case, the total number of infected cases could be reduced by more than 90%. Delay in implementing quarantine measures would drastically reduce their effectiveness.