Abstract: Nonlinear dynamics of interactions between clusters of neurons via complex networks lie at the base of all brain activity. How such communication between brain regions gives rise to the rich behavioral repertoire of the organism has been a long-standing question. In this talk, we will explore this question by looking at the simulations of collective dynamics of a detailed network of cortical areas in the Macaque brain recently compiled from the CoCoMac database, as well as, a model of global coupled brain regions used as a benchamrk. To understand the large-scale dynamics of the brain, we simulate it at the mesoscopic level with each node representing a local region of cortex, comprising between 10^3-10^6 neurons. The dynamical behavior of each such region has been described using a phenomenological model consisting of a pool of excitatory neurons coupled to a pool of inhibitory neurons, which exhibits oscillations over a large range of parameter values. Coupling these regions according to the Macaque cortical network produces activation patterns strikingly similar to those observed in recordings from the brain. Our results help to connect recent experimental findings of the olfactory system and suggest that a part of the complicated activity patterns seen in the brain may be explained even without a full knowledge of its wiring diagram.

Abstract: Betweenness is an important network property to study system performance against perturbations (such as random natural disasters and intended terrorism attacks). Traditionally, betweenness of a node/link is defined as how many times the node/link appears as intermediate node/link in all shortest paths between nodes. Traditional betweenness can help to answer how likely the 1st best paths between nodes will be cut off by perturbations. However, in reality, it is more concerned with a general situation, i.e., how likely those paths whose lengths are within a given range will be affected by perturbations. For instance, for a researcher to attend a conference, whether the 1st best path is available is not important at all, and instead, whether s/he can arrive on time is the key. Unfortunately, this general situation has barely been discussed in literature. To assess network performance in the general situation, we propose a generalized betweenness which is mathematically defined as how many times a node/link appears as intermediate node/link in all those paths whose lengths are within a given range. No existing method can effectively calculate the generalized betweenness, because it is difficult to find out, between every pair of nodes in a network, all those paths whose lengths are within a given range. By modifying a newly reported nature-inspired method, i.e., ripple-spreading algorithm (RSA), it becomes possible to calculate the generalized betweenness. Surprisingly, the proposed RSA can effectively find out all those paths whose lengths are within a given range by just a single run of ripple relay race. This work makes progress towards the general performance assessment of a network system against perturbations.

Abstract: Urban street structure is a snapshot of the human mobility and an important medium facilitating the human interaction. Previous studies have analyzed the topology and morphology of street structure in various ways; fractal patterns, complex spatial network and so on. In practical term, it is also important to discuss how street networks are used by people. There are studies analyzing the efficiency, accessibility and road usage in the street networks too. In those studies, people usually investigate routes, either empirical routes or theoretical routes, to understand the functionality of the network. The travel route is a good proxy to understand the street structure and city attributes from user perspective. It is a selected path from the whole network by people or under the given standard, so it reveals how people use the networks. The selected route is also influenced by various factors such as spatial pattern and travel demand of city. Thus studying the empirical or theoretical optimal routes can help us understand the urban characteristics which are often hidden. For instance, fastest routes show the distribution of traffic in a city as well as the street structure. In this paper, we analyze the geometric property of routes to understand the geometry of practical street network where hierarchy and traffic is included. We use the two types of optimal routes collected from time minimizing algorithm and distance minimizing algorithm via the OpenStreetMap API. We suggest a new metric, center-philic level, to measure how much a route is bent toward the city center. We measured the center-philic level of a number of route within 30km radius from the center. By analyzing the center-philic level for different location of routes, we can understand and simplify the geometry of street network based on the center. The center-philic level patterns for two different algorithms can also reveal the effect of street hierarchy and traffic. In urban transportation, we can imagine two forces competing each other. By the agglomeration of business and people inside of city, street networks become denser around the center area to satisfy the demand. Such centralized street networks attract traffic toward inside of city. However many cities have arterial roads located outside of city to disperses the traffic concentrated on the inside of city. The arterial roads act as another force pulling traffic toward outside of the city. This tendency is well captured by our suggested metric. We firstly compare the general average center-philic level of both shortest and fastest routes to point out the fundamental difference between them. Later we analyze the center-philic level of individual cities and discuss how the metric can explain the street layout and street hierarchy.

Abstract: Network analysis has become a powerful tool to analyze complex systems over wide range of topics. For last two decades, researchers have made much progress particularly in the topics of disease spreading, social interaction, biological metabolism, neural network, urban mobility, etc. However, energy system has yet been plentifully covered. In this talk, we seek to apply network theory into electric power systems. Firstly, we integrate network analysis into environmental impact analysis. We introduce energy distance in order to estimate the greenhouse gas emissions of electricity transmission taking both the amount of electricity consumption and transmission distance into account. Secondly, we analyze the functional stability of power grids. The stable synchronization of power-grid nodes is the essential condition for the secure electric power systems. We investigate the transition of the synchronization stability of power-grid nodes and classify nodes based on the transition patterns. We conclude that network analysis is a good complement for energy system analysis.

Abstract: Scientific collaboration plays an important role in the knowledge production and scientific development. The researchers have constructed several network models of scientific collaboration. In traditional collaboration network, two scientists are linked if they have coauthored one paper. However, this construction of network undervalues the role of the first author. In this paper, we propose a new collaboration network model considering the importance of the partnership between the first author and others. We make an empirical analysis based on the data of American Physical Society (APS). The?results?show?that?there?are??some? differences?of properties?between?the?new?network?and?the?traditional?one. And the node importance is studied on the new network to identify potential researchers.

Abstract: Nuclear fusion research, which originated from atomic weapon developments by USA and USSR, attracts public attention as a promising energy source for the future. After the Cold War, Nations have collaborated in order to build research capacity in nuclear fusion. ITER (International Thermonuclear Experimental Reactor) is an example of `big science' projects at the international level. Scientists from different disciplines involve in the project. The goal of our study is to investigate collaboration structure of nuclear fusion research and its evolution through an open access bibliometric database, Microsoft Academic Graph (MAG). We examine not only scientific journal citations but also the impact of IAEA Fusion Energy Conference on the research field. Dynamics of co-authorship networks reveal how nations take part and collaborate in nuclear fusion research. We expect that this study would be helpful for managing research activities and for suggesting national S&T policies.

Abstract: In the real world, infrastructure networks such as communication networks, power grid networks, transportation networks, solidly underlie the development of the whole society. The structure of infrastructure networks become more and more complicate and always couple together to perform intact service capacity. There commonly exist dependency among components as well as sub-networks, which make failure propagation. Currently, numerous literatures focused on the vulnerability and robustness of classical complex networks (e.g. random network, regular network, small-world network and scale-free network) under malicious attacks or random attacks. As one kind of real-world networks, besides having the topological characteristic of classical complex network, infrastructure networks are restricted by social-economic and geographical factors, so that they have short length links and some of them are planar graphs. Infrastructure network distributes in a specific spatially geographical domain, which probably exposures to real-world localized attacks (such as natural disasters). Recently, investigation on spatially embedded infrastructure networks under localized attacks is getting more and more attention. But full consideration of spatial characteristic of nodes and links in the robustness investigation of infrastructure network is still a big challenge due to involving real-world factors. In this paper, we aim to study the robustness of spatially embedded and coupled infrastructure networks under localized attacks. We first generate different kinds of spatially embedded infrastructure networks. Then a density-based index is proposed to depict the spatial characteristic of infrastructure network, dependency links among sub-networks are placed according to geographical restriction. Localized attacks are described by the circles with different radius. Finally, numerical simulation is conducted and the result illustrates that the spatial characteristic of infrastructure network and location of dependency links have significant effect on the robustness of infrastructure network under localized attack.

Abstract: The study of human mobility patterns is of both theoretical and practical values in many aspects. For long-distance travels, a few research endeavors have shown that the displacements of human travels follow the power-law distribution. However, the intra-urban travels do not simply follow the same power-law of longer-distance travels. What?s more, controversies remain in the issue of the scaling law of human mobility in intra-urban areas. In this work we focus on the mobility pattern of taxi passengers by examining five datasets of the three metropolitans of New York, Dalian and Nanjing. Through statistical analysis, we find the mixed distribution of lognormal and power-law better explain both the displacement and the duration time of taxi trips, as well as the vacant time of taxicabs, in all the examined cities. The universality of scaling law of human mobility is subsequently discussed, in accordance with the data analytics.

Abstract: We live in a society, where people interact with each other. An individual's opinion is formed with a neighbor's opinion. How do people accept a different opinion? Threshold model is the most well-known and proved theoretical model for the question. The model assumes that a person receives a different opinion when the ratio of their neighbor, whose opinion is different with one's opinion, is higher than one's threshold for an acceptance. Much research has been conducted on the opinion dynamics in the society with 'threshold model', and diverse aspects of key features in opinion dynamics have been revealed by the model. Most of the researches related with the threshold model in opinion dynamics are placed in the homogeneous threshold assumption. It means every person has the same level of threshold for an acceptance of an opinion. The usefulness of this assumption is clear in terms of statistical analysis. Howeve, we all know that heterogenuity of threshold exists. Even the threshold has a relationship with social capital like the number of neighbors, which can be represented as the characteristic of a network structure. Surprisingly, investigation on the correlation between network structure and threshold has not been conducted well. In this study, we investigate the influence of the correlation between the number of neighbors (influential level) and individual threshold on opinion dynamics. Given the scale-free network, as the representative network model of society, we change a level of correlation ($\beta$) from negative (-1) to positive (+1): 'negative' ('positive') is for the case that small (large) degree node has high threshold. The minimum of a threshold is set to 0.5 since we assume that people change their opinion when there are people more than a half of their neighbor generally. Additionally, opinion is composed with 0 and 1, and 60% of people in the network has opinion 1 as an initial condition. In this setting, we conduct the opinion dynamics while changing the correlation with degree and threshold. We found that the importance of the correlation betwen degree and threshold as to opinion dynamics. In this result, negative correlation region spreads information to entire system, but the positive correlation has a finite steady state. Additionally, there is a transition point around 0.5 regarding to correlation level. So far, we are conducting a finite size scaling analysis to figure out the characteristic of the transition, however, we were able to deduce from this result, there could be a transition of the system. To see the specific origin of this asymmetric contagion, we measured fixation time and final opinion. Fixation time is the time the node is fixed its final opinion. Moreover, the tendency of fixation time changes in positive correlation around the critical point, which is close to 0.5. This point is very close with the transition point for opinion switching. It means there is a balanced dependency between influential level and threshold so that from the point, the system can be divided with two different regimes: one is the regime for threshold dominant effect, the other is for the confused information effect. Another interesting result is: low influential nodes start to fluctuate more with high degree of dependency. Even though they have low threshold with positive correlation, they still receive a lot of influences from high degree nodes so they confused with mixed information from high degree nodes. This result suggests that the correlation between network structure and threshold of an acceptance can be important for opinion formation. Since we often experience the effect of correlation between influential level and threshold, it is worth to stress out the meaning of result. We can interpret the result as follows: positive correlation with influential level and conservative characteristic can block the unification of opinion in less influential group so that the majority opinion could not reach the whole society.

Abstract: Inferring the model of animal movements and aggregation have been a long-term challenging task. Although numerous of realistic-looking models have been proposed, model-based methods often rely on untested assumptions. Besides, many sets of microscopic hypotheses can produce the same macroscopic behaviors, it is dubious that they uncovered the real inherent mechanism. In this work, we conducted experiments with ants in two-dimensional circular surface. We try to infer behavioral rules directly from experimental data instead of traditional model-based research strategy. By defining a new metric to measure the ants aggregation extent, we study the influence of edge ?wall? and other individual on ants aggregation. Instead of studying large groups from 50 to 150, we study small number of ants from 1 to 3. By analyzing and comparing the data, we proposed a simple yet effective model, which may help to account for the micro-foundation of ants aggregation and infer how they amplify aggregation extent in large groups.

Abstract: As the increasing interest in the investigation of collective motion of a group of animals, it is important to tracking multiple moving animals and acquiring their position over time and space. There are several studies that have tried to solve this problem and make this data acquisition automated. However, none of these studies has solved the problem very well and automatically tracking is very difficult thanks to the individual?s various shape, complex motion and frequent occlusion. There are several published algorithms working on this problem, usually aim at one special specie, zebra fish, for instance. However, these algorithms have very high demands on the video quality, such as high frame rates, high image resolution and steady background, some of them are very time-consuming. Here we have developed an integrated approach based on artificial neural networks that enables us to automatically extract individuals? trajectories from both high quality and low quality videos. First we combine a background subtraction method and artificial neural networks to effectively detect the individuals. Then we use a linear assignment model to track the individuals. At last, we build a function to measure the confidence coefficient for each frame to help correct the possible errors. We applied our method to track different fish videos, the results showed that our method has a high efficiency and accuracy in most situations.

Abstract: Collective motion of fish is an interesting research field. There is an essential question that how to quantify the how collective a group is, namely, how to recognize that fish in the group are interacting. Generally, researchers quantify the strength of interaction between fish intuitively using the correlation of velocities or spins, and measure the whole group?s polarization and rotation, etc. to determine whether it?s synergetic. It is because researchers consider the animal group motions as multi-body physical phenomena. However, it poses a problem that if an animal group does not display a visible collective structure, these physical statistics will fail to recognize the underlying mechanism. We introduce a new measure based on differential mutual information from information theory to quantify the co-ordination. Information theory is used in many fields but rarely in the field of collective animal motion. The original mutual information is a value to measure the correspondence of two signals, if there is any correspondence between these two agents? moving, mutual information will reveal it, while old statistics like polarization fail if the relationship is strange. The new statistic will be compared with classical statistics like polarization and correlation on Couzin model and Vicsek model. We will try different parameters to analyze the properties of all these statistics. And we will show that the new statistic is more efficient than old statistics when the model displays confusion and has even efficiency when the model displays order. We record the trajectories of different amounts of Glass Goldfishes swimming in a tank with a radius of 40 centimeters. The new statistic is used to measure the co-ordination of the groups. We will show that the co-ordination grows while the fish amount grows. In short, we will illustrate that the new measure can be helpful to reveal the strength of the co-ordination in a group.