Classification of major depressive disorder using an attention-guided unified deep convolutional neural network and individual structural covariance network.

Major depressive disorder (MDD) is the second leading cause of disability worldwide. Currently, the structural magnetic resonance imaging-based MDD diagnosis models mainly utilize local grayscale information or morphological characteristics in a single site with small samples. Emerging evidence has demonstrated that different brain structures in different circuits have distinct developmental timing, but mature coordinately within the same functional circuit. Thus, establishing an attention-guided unified classification framework with deep learning and individual structural covariance networks in a large multisite dataset could facilitate developing an accurate diagnosis strategy. Our results showed that attention-guided classification could improve the classification accuracy from primary 75.1% to ultimate 76.54%. Furthermore, the discriminative features of regional covariance connectivities and local structural characteristics were found to be mainly located in prefrontal cortex, insula, superior temporal cortex, and cingulate cortex, which have been widely reported to be closely associated with depression. Our study demonstrated that our attention-guided unified deep learning framework may be an effective tool for MDD diagnosis. The identified covariance connectivities and structural features may serve as biomarkers for MDD.

Epidemic spreading on multi-layer networks with active nodes.

Investigations on spreading dynamics based on complex networks have received widespread attention these years due to the COVID-19 epidemic, which are conducive to corresponding prevention policies. As for the COVID-19 epidemic itself, the latent time and mobile crowds are two important and inescapable factors that contribute to the significant prevalence. Focusing on these two factors, this paper systematically investigates the epidemic spreading in multiple spaces with mobile crowds. Specifically, we propose a SEIS (Susceptible-Exposed-Infected-Susceptible) model that considers the latent time based on a multi-layer network with active nodes which indicate the mobile crowds. The steady-state equations and epidemic threshold of the SEIS model are deduced and discussed. And by comprehensively discussing the key model parameters, we find that (1) due to the latent time, there is a "cumulative effect" on the infected, leading to the "peaks" or "shoulders" of the curves of the infected individuals, and the system can switch among three states with the relative parameter combinations changing; (2) the minimal mobile crowds can also cause the significant prevalence of the epidemic at the steady state, which is suggested by the zero-point phase change in the proportional curves of infected individuals. These results can provide a theoretical basis for formulating epidemic prevention policies.

Minimum taxi fleet algorithm considering human spatiotemporal behaviors.

With the development of information technology, more and more travel data have provided great convenience for scholars to study the travel behavior of users. Planning user travel has increasingly attracted researchers' attention due to its great theoretical significance and practical value. In this study, we not only consider the minimum fleet size required to meet the urban travel needs but also consider the travel time and distance of the fleet. Based on the above reasons, we propose a travel scheduling solution that comprehensively considers time and space costs, namely, the Spatial-Temporal Hopcroft-Karp (STHK) algorithm. The analysis results show that the STHK algorithm not only significantly reduces the off-load time and off-load distance of the fleet travel by as much as 81% and 58% and retains the heterogeneous characteristics of human travel behavior. Our study indicates that the new planning algorithm provides the size of the fleet to meet the needs of urban travel and reduces the extra travel time and distance, thereby reducing energy consumption and reducing carbon dioxide emissions. Concurrently, the travel planning results also conform to the basic characteristics of human travel and have important theoretical significance and practical application value.

Identifying Important Nodes in Trip Networks and Investigating Their Determinants.

Describing travel patterns and identifying significant locations is a crucial area of research in transportation geography and social dynamics. Our study aims to contribute to this field by analyzing taxi trip data from Chengdu and New York City. Specifically, we investigate the probability density distribution of trip distance in each city, which enables us to construct long- and short-distance trip networks. To identify critical nodes within these networks, we employ the PageRank algorithm and categorize them using centrality and participation indices. Furthermore, we explore the factors that contribute to their influence and observe a clear hierarchical multi-centre structure in Chengdu's trip networks, while no such phenomenon is evident in New York City's. Our study provides insight into the impact of trip distance on important nodes within trip networks in both cities and serves as a reference for distinguishing between long and short taxi trips. Our findings also reveal substantial differences in network structures between the two cities, highlighting the nuanced relationship between network structure and socio-economic factors. Ultimately, our research sheds light on the underlying mechanisms shaping transportation networks in urban areas and offers valuable insights into urban planning and policy making.

Effects of the information-driven awareness on epidemic spreading on multiplex networks.

In this study, we examine the impact of information-driven awareness on the spread of an epidemic from the perspective of resource allocation by comprehensively considering a series of realistic scenarios. A coupled awareness-resource-epidemic model on top of multiplex networks is proposed, and a Microscopic Markov Chain Approach is adopted to study the complex interplay among the processes. Through theoretical analysis, the infection density of the epidemic is predicted precisely, and an approximate epidemic threshold is derived. Combining both numerical calculations and extensive Monte Carlo simulations, the following conclusions are obtained. First, during a pandemic, the more active the resource support between individuals, the more effectively the disease can be controlled; that is, there is a smaller infection density and a larger epidemic threshold. Second, the disease can be better suppressed when individuals with small degrees are preferentially protected. In addition, there is a critical parameter of contact preference at which the effectiveness of disease control is the worst. Third, the inter-layer degree correlation has a "double-edged sword" effect on spreading dynamics. In other words, when there is a relatively lower infection rate, the epidemic threshold can be raised by increasing the positive correlation. By contrast, the infection density can be reduced by increasing the negative correlation. Finally, the infection density decreases when raising the relative weight of the global information, which indicates that global information about the epidemic state is more efficient for disease control than local information.

Application of network link prediction in drug discovery.

BACKGROUND: Technological and research advances have produced large volumes of biomedical data. When represented as a network (graph), these data become useful for modeling entities and interactions in biological and similar complex systems. In the field of network biology and network medicine, there is a particular interest in predicting results from drug-drug, drug-disease, and protein-protein interactions to advance the speed of drug discovery. Existing data and modern computational methods allow to identify potentially beneficial and harmful interactions, and therefore, narrow drug trials ahead of actual clinical trials. Such automated data-driven investigation relies on machine learning techniques. However, traditional machine learning approaches require extensive preprocessing of the data that makes them impractical for large datasets. This study presents wide range of machine learning methods for predicting outcomes from biomedical interactions and evaluates the performance of the traditional methods with more recent network-based approaches. RESULTS: We applied a wide range of 32 different network-based machine learning models to five commonly available biomedical datasets, and evaluated their performance based on three important evaluations metrics namely AUROC, AUPR, and F1-score. We achieved this by converting link prediction problem as binary classification problem. In order to achieve this we have considered the existing links as positive example and randomly sampled negative examples from non-existant set. After experimental evaluation we found that Prone, ACT and [Formula: see text] are the top 3 best performers on all five datasets. CONCLUSIONS: This work presents a comparative evaluation of network-based machine learning algorithms for predicting network links, with applications in the prediction of drug-target and drug-drug interactions, and applied well known network-based machine learning methods. Our work is helpful in guiding researchers in the appropriate selection of machine learning methods for pharmaceutical tasks.

Simulating two-phase taxi service process by random walk theory.

City taxi service systems have been empirically studied by a number of data-driven methods. However, their underlying mechanisms are hard to understand because the present mathematical models neglect to explain a (whole) taxi service process that includes a pair of on-load phase and off-load phase. In this paper, by analyzing a large amount of taxi servicing data from a large city in China, we observe that the taxi service process shows different temporal and spatial features according to the on-load phase and off-load phase. Moreover, our correlation analysis results demonstrate the lack of dependence between the on-load phase and the off-load phase. Hence, we introduce two independent random walk models based on the Langevin equation to describe the underlying mechanism and to understand the temporal and spatial features of the taxi service process. Our study attempts to formulate the mathematical framework for simulating the taxi service process and better understanding of its underlying mechanism.

Effects of heterogeneous self-protection awareness on resource-epidemic coevolution dynamics.

Recent studies have demonstrated that the allocation of individual resources has a significant influence on the dynamics of epidemic spreading. In the real scenario, individuals have a different level of awareness for self-protection when facing the outbreak of an epidemic. To investigate the effects of the heterogeneous self-awareness distribution on the epidemic dynamics, we propose a resource-epidemic coevolution model in this paper. We first study the effects of the heterogeneous distributions of node degree and self-awareness on the epidemic dynamics on artificial networks. Through extensive simulations, we find that the heterogeneity of self-awareness distribution suppresses the outbreak of an epidemic, and the heterogeneity of degree distribution enhances the epidemic spreading. Next, we study how the correlation between node degree and self-awareness affects the epidemic dynamics. The results reveal that when the correlation is positive, the heterogeneity of self-awareness restrains the epidemic spreading. While, when there is a significant negative correlation, strong heterogeneous or strong homogeneous distribution of the self-awareness is not conducive for disease suppression. We find an optimal heterogeneity of self-awareness, at which the disease can be suppressed to the most extent. Further research shows that the epidemic threshold increases monotonously when the correlation changes from most negative to most positive, and a critical value of the correlation coefficient is found. When the coefficient is below the critical value, an optimal heterogeneity of self-awareness exists; otherwise, the epidemic threshold decreases monotonously with the decline of the self-awareness heterogeneity. At last, we verify the results on four typical real-world networks and find that the results on the real-world networks are consistent with those on the artificial network.

Optimizing spreading dynamics in interconnected networks.

Adding edges between layers of interconnected networks is an important way to optimize the spreading dynamics. While previous studies mostly focused on the case of adding a single edge, the theoretical optimal strategy for adding multiple edges still need to be studied. In this study, based on the susceptible-infected-susceptible model, we investigate the problem of maximizing the stationary spreading prevalence in interconnected networks. For two isolated networks, we maximize the spreading prevalence near the critical point by choosing multiple interconnecting edges. We present a theoretical analysis based on the discrete-time Markov chain approach to derive the approximate optimal strategy. The optimal interlayer structure predicted by the strategy maximizes the spreading prevalence, meanwhile minimizing the spreading outbreak threshold for the interconnected network simultaneously. Numerical simulations on synthetic and real-world networks show that near the critical point, the proposed strategy gives better performance than connecting large degree nodes and randomly connecting.

Repulsive synchronization in complex networks.

Synchronization in complex networks characterizes what happens when an ensemble of oscillators in a complex autonomous system become phase-locked. We study the Kuramoto model with a tunable phase-lag parameter α in the coupling term to determine how phase shifts influence the synchronization transition. The simulation results show that the phase frustration parameter leads to desynchronization. We find two global synchronization regions for α∈[0,2π) when the coupling is sufficiently large and detect a relatively rare network synchronization pattern in the frustration parameter near α=π. We call this frequency-locking configuration as "repulsive synchronization," because it is induced by repulsive coupling. Since the repulsive synchronization cannot be described by the usual order parameter r, the parameter frequency dispersion is introduced to detect synchronization.

Self-awareness control effect of cooperative epidemics on complex networks.

Coinfection mechanism is a common interacting mode between multiple diseases in real spreading processes, where the diseases mutually increase their susceptibility, and has aroused widespread studies in network science. We use the bond percolation theory to characterize the coinfection model under two self-awareness control strategies, including immunization strategy and quarantine strategy, and to study the impacts of the synergy effect and control strategies on cooperative epidemics. We find that strengthening the synergy effect can reduce the epidemic threshold and enhance the outbreak size of coinfected networks. On Erdos-Rényi networks, the synergy effect will induce a crossover phenomenon of phase transition, i.e., make the type of phase transition from being continuous to discontinuous. Self-awareness control strategies play a non-negligible role in suppressing cooperative epidemics. In particular, increasing immunization or the quarantine rate can enhance the epidemic threshold and reduce the outbreak size of cooperative epidemics, and lead to a crossover phenomenon of transition from being discontinuous to continuous. The impact of quarantine strategy on cooperative epidemics is more significant than the immunization strategy, which is verified on scale-free networks.

Accurate detection of hierarchical communities in complex networks based on nonlinear dynamical evolution.

One of the most challenging problems in network science is to accurately detect communities at distinct hierarchical scales. Most existing methods are based on structural analysis and manipulation, which are NP-hard. We articulate an alternative, dynamical evolution-based approach to the problem. The basic principle is to computationally implement a nonlinear dynamical process on all nodes in the network with a general coupling scheme, creating a networked dynamical system. Under a proper system setting and with an adjustable control parameter, the community structure of the network would "come out" or emerge naturally from the dynamical evolution of the system. As the control parameter is systematically varied, the community hierarchies at different scales can be revealed. As a concrete example of this general principle, we exploit clustered synchronization as a dynamical mechanism through which the hierarchical community structure can be uncovered. In particular, for quite arbitrary choices of the nonlinear nodal dynamics and coupling scheme, decreasing the coupling parameter from the global synchronization regime, in which the dynamical states of all nodes are perfectly synchronized, can lead to a weaker type of synchronization organized as clusters. We demonstrate the existence of optimal choices of the coupling parameter for which the synchronization clusters encode accurate information about the hierarchical community structure of the network. We test and validate our method using a standard class of benchmark modular networks with two distinct hierarchies of communities and a number of empirical networks arising from the real world. Our method is computationally extremely efficient, eliminating completely the NP-hard difficulty associated with previous methods. The basic principle of exploiting dynamical evolution to uncover hidden community organizations at different scales represents a "game-change" type of approach to addressing the problem of community detection in complex networks.

Dual-induced multifractality in online viewing activity.

Although recent studies have found that the long-term correlations relating to the fat-tailed distribution of inter-event times exist in human activity and that these correlations indicate the presence of fractality, the property of fractality and its origin have not been analyzed. We use both detrended fluctuation analysis and multifractal detrended fluctuation analysis to analyze the time series in online viewing activity separating from Movielens and Netflix. We find long-term correlations at both the individual and communal levels and that the extent of correlation at the individual level is determined by the activity level. These long-term correlations also indicate that there is fractality in the pattern of online viewing. We first find a multifractality that results from the combined effect of the fat-tailed distribution of inter-event times (i.e., the times between successive viewing actions of individuals) and the long-term correlations in online viewing activity and verify this finding using three synthesized series. Therefore, it can be concluded that the multifractality in online viewing activity is caused by both the fat-tailed distribution of inter-event times and the long-term correlations and that this enlarges the generic property of human activity to include not just physical space but also cyberspace.

Controlling epidemic outbreak based on local dynamic infectiousness on complex networks.

Resources are limited in epidemic containment; how to optimally allocate the limited resources in suppressing the epidemic spreading has been a challenging problem. To find an effective resource allocation strategy, we take the infectiousness of each infected node into consideration. By studying the interplay between the resource allocation and epidemic spreading, we find that the spreading dynamics of epidemic is affected by the preferential resource allocation. There are double phase transitions of the fraction of infected nodes, which are different from the classical epidemic model. More importantly, we find that the preferential resource allocation has double-edged sword effects on the disease spreading. When there is a small transmission rate, the infected fraction at the steady state decreases with the increment of degree of resource allocation preference, which indicates that resources of the healthy nodes should be allocated preferentially to the high infectious nodes to constrain the disease spreading. Moreover, when there is a large transmission rate, the fraction of infected nodes at the steady state increases with the increment of the degree of the preference, but the resource allocation is determined by the stage of epidemic spreading. Namely, in the early stage of the disease spreading, resources should be allocated preferentially to the high infectious nodes similar to the case of a small transmission rate. While after the early stage, resources should be allocated to the low infectious nodes. Based on the findings, we propose a simple resource allocation strategy that can adaptively change with the current fraction of infected nodes and the disease can be suppressed to the most extent under the proposed strategy.

Non-Markovian character in human mobility: Online and offline.

The dynamics of human mobility characterizes the trajectories that humans follow during their daily activities and is the foundation of processes from epidemic spreading to traffic prediction and information recommendation. In this paper, we investigate a massive data set of human activity, including both online behavior of browsing websites and offline one of visiting towers based mobile terminations. The non-Markovian character observed from both online and offline cases is suggested by the scaling law in the distribution of dwelling time at individual and collective levels, respectively. Furthermore, we argue that the lower entropy and higher predictability in human mobility for both online and offline cases may originate from this non-Markovian character. However, the distributions of individual entropy and predictability show the different degrees of non-Markovian character between online and offline cases. To account for non-Markovian character in human mobility, we apply a protype model with three basic ingredients, namely, preferential return, inertial effect, and exploration to reproduce the dynamic process of online and offline human mobilities. The simulations show that the model has an ability to obtain characters much closer to empirical observations.

Community structure revealed by phase locking.

Community structure can naturally emerge in paths to synchronization, and scratching it from the paths is a tough issue that accounts for the diverse dynamics of synchronization. In this paper, with assumption that the synchronization on complex networks is made up of local and collective processes, we proposed a scheme to lock the local synchronization (phase locking) at a stable state, meanwhile, suppress the collective synchronization based on Kuramoto model. Through this scheme, the network dynamics only contains the local synchronization, which suggests that the nodes in the same community synchronize together and these synchronization clusters well reveal the community structure of network. Furthermore, by analyzing the paths to synchronization, the relations or overlaps among different communities are also obtained. Thus, the community detection based on the scheme is performed on five real networks and the observed community structures are much more apparent than modularity-based fast algorithm. Our results not only provide a deep insight to understand the synchronization dynamics on complex network but also enlarge the research scope of community detection.

Synchronization-based approach for detecting functional activation of brain.

In this paper, we investigate a synchronization-based, data-driven clustering approach for the analysis of functional magnetic resonance imaging (fMRI) data, and specifically for detecting functional activation from fMRI data. We first define a new measure of similarity between all pairs of data points (i.e., time series of voxels) integrating both complete phase synchronization and amplitude correlation. These pairwise similarities are taken as the coupling between a set of Kuramoto oscillators, which in turn evolve according to a nearest-neighbor rule. As the network evolves, similar data points naturally synchronize with each other, and distinct clusters will emerge. The clustering behavior of the interaction network of the coupled oscillators, therefore, mirrors the clustering property of the original multiple time series. The clustered regions whose cross-correlation coefficients are much greater than other regions are considered as the functionally activated brain regions. The analysis of fMRI data in auditory and visual areas shows that the recognized brain functional activations are in complete correspondence with those from the general linear model of statistical parametric mapping, but with a significantly lower time complexity. We further compare our results with those from traditional K-means approach, and find that our new clustering approach can distinguish between different response patterns more accurately and efficiently than the K-means approach, and therefore more suitable in detecting functional activation from event-related experimental fMRI data.

A competition model for prediction of admission scores of colleges and universities in Chinese college entrance examination.

Predicting the admission scores of colleges and universities is significant for high school graduates in the College Entrance Examination in China (which is also called "Gaokao" for short). The practice of parallel application for the students after Gaokao not only puts forward a question about how students could make the best of their scores and make the best choice, but also results in the strong competition among different colleges and universities, with the institutions all striving to admit high-performing students in this examination. However, existing prevailing prediction algorithms and models of the admission score of the colleges and universities based on machine learning methods do not take such competitive relationship into consideration, but simply make predictions for individual college or university, causing low predication accuracy and poor generalization capability. This paper intends to analyze such competitive relationship by extracting the important features (e.g., project, location and score discrepancy) of colleges and universities. A novel competition model incorporating the coarse clustering is thus proposed to make the predictions for colleges and universities in a same cluster. By using Gaokao data of Shanxi province in China from 2016 to 2019, we testify the proposed model in comparison with several benchmark methods. The experimental results show that the precision within the error of 3 points and 5 points are 7.3% and 2.8% higher respectively than the second-best algorithm. It has proven that the competition model has the capability to fit the competitive relationship, thus improving the predication accuracy to a large extent. Theoretically, the method proposed could provide a more advanced and comprehensive view about the analysis of factors that may influence the admission score of higher institutions. Practically, the model proposed with high accuracy could help the students make the best of their scores and apply for the college and universities more scientifically.

Multisite Autism Spectrum Disorder Classification Using Convolutional Neural Network Classifier and Individual Morphological Brain Networks.

Autism spectrum disorder (ASD) is a range of neurodevelopmental disorders with behavioral and cognitive impairment and brings huge burdens to the patients' families and the society. To accurately identify patients with ASD from typical controls is important for early detection and early intervention. However, almost all the current existing classification methods for ASD based on structural MRI (sMRI) mainly utilize the independent local morphological features and do not consider the covariance patterns of these features between regions. In this study, by combining the convolutional neural network (CNN) and individual structural covariance network, we proposed a new framework to classify ASD patients with sMRI data from the ABIDE consortium. Moreover, gradient-weighted class activation mapping (Grad-CAM) was applied to characterize the weight of features contributing to the classification. The experimental results showed that our proposed method outperforms the currently used methods for classifying ASD patients with the ABIDE data and achieves a high classification accuracy of 71.8% across different sites. Furthermore, the discriminative features were found to be mainly located in the prefrontal cortex and cerebellum, which may be the early biomarkers for the diagnosis of ASD. Our study demonstrated that CNN is an effective tool to build the framework for the diagnosis of ASD with individual structural covariance brain network.

Equivalence and its invalidation between non-Markovian and Markovian spreading dynamics on complex networks.

Epidemic spreading processes in the real world depend on human behaviors and, consequently, are typically non-Markovian in that the key events underlying the spreading dynamics cannot be described as a Poisson random process and the corresponding event time is not exponentially distributed. In contrast to Markovian type of spreading dynamics for which mathematical theories have been well developed, we lack a comprehensive framework to analyze and fully understand non-Markovian spreading processes. Here we develop a mean-field theory to address this challenge, and demonstrate that the theory enables accurate prediction of both the transient phase and the steady states of non-Markovian susceptible-infected-susceptible spreading dynamics on synthetic and empirical networks. We further find that the existence of equivalence between non-Markovian and Markovian spreading depends on a specific edge activation mechanism. In particular, when temporal correlations are absent on active edges, the equivalence can be expected; otherwise, an exact equivalence no longer holds.