Search

In this paper a weighted complex network is used to detect communities of basketball players on the basis of their performances. A sparsification procedure to remove weak edges is also applied. In our proposal, at each removal of an edge the best community structure of the "giant component" is calculated, maximizing the modularity as a measure of compactness within communities and separation among communities. The "sparsification transition" is confirmed by the normalized mutual information. In this way, not only the best distribution of nodes into communities is found, but also the ideal number of communities as well. An application to community detection of basketball players for the NBA regular season 2020–2021 is presented. The proposed methodology allows a data driven decision making process in basketball. [ABSTRACT FROM AUTHOR]

In recent years, transformer-based models have emerged as powerful tools for natural language processing tasks, demonstrating remarkable performance in several domains. However, they still present significant limitations. These shortcomings become more noticeable when dealing with highly specific and complex concepts, particularly within the scientific domain. For example, transformer models have particular difficulties when processing scientific articles due to the domain-specific terminologies and sophisticated ideas often encountered in scientific literature. To overcome these challenges and further enhance the effectiveness of transformers in specific fields, researchers have turned their attention to the concept of knowledge injection. Knowledge injection is the process of incorporating outside knowledge into transformer models to improve their performance on certain tasks. In this paper, we present a comprehensive study of knowledge injection strategies for transformers within the scientific domain. Specifically, we provide a detailed overview and comparative assessment of four primary methodologies, evaluating their efficacy in the task of classifying scientific articles. For this purpose, we constructed a new benchmark including both 24K labelled papers and a knowledge graph of 9.2K triples describing pertinent research topics. We also developed a full codebase to easily re-implement all knowledge injection strategies in different domains. A formal evaluation indicates that the majority of the proposed knowledge injection methodologies significantly outperform the baseline established by Bidirectional Encoder Representations from Transformers. [ABSTRACT FROM AUTHOR]

In recent years, transformer-based models have emerged as powerful tools for natural language processing tasks, demonstrating remarkable performance in several domains. However, they still present significant limitations. These shortcomings become more noticeable when dealing with highly specific and complex concepts, particularly within the scientific domain. For example, transformer models have particular difficulties when processing scientific articles due to the domain-specific terminologies and sophisticated ideas often encountered in scientific literature. To overcome these challenges and further enhance the effectiveness of transformers in specific fields, researchers have turned their attention to the concept of knowledge injection. Knowledge injection is the process of incorporating outside knowledge into transformer models to improve their performance on certain tasks. In this paper, we present a comprehensive study of knowledge injection strategies for transformers within the scientific domain. Specifically, we provide a detailed overview and comparative assessment of four primary methodologies, evaluating their efficacy in the task of classifying scientific articles. For this purpose, we constructed a new benchmark including both 24K labelled papers and a knowledge graph of 9.2K triples describing pertinent research topics. We also developed a full codebase to easily re-implement all knowledge injection strategies in different domains. A formal evaluation indicates that the majority of the proposed knowledge injection methodologies significantly outperform the baseline established by Bidirectional Encoder Representations from Transformers. [ABSTRACT FROM AUTHOR]

In this work, we analyze the hyperbolicity of real-world networks, a geometric quantity that measures if a space is negatively curved. We provide two improvements in our understanding of this quantity: first of all, in our interpretation, a hyperbolic network is "aristocratic", since few elements "connect" the system, while a non-hyperbolic network has a more "democratic" structure with a larger number of crucial elements. The second contribution is the introduction of the average hyperbolicity of the neighbors of a given node. Through this definition, we outline an "influence area" for the vertices in the graph. We show that in real networks the influence area of the highest degree vertex is small in what we define "local" networks (i.e., social or peer-to-peer networks), and large in "global" networks (i.e., power grid, metabolic networks, or autonomous system networks). [ABSTRACT FROM AUTHOR]

In this work we are interested in identifying clusters of “positional equivalent” actors, i.e. actors who play a similar role in a system. In particular, we analyze weighted bipartite networks that describes the relationships between actors on one side and features or traits on the other, together with the intensity level to which actors show their features. We develop a methodological approach that takes into account the underlying multivariate dependence among groups of actors. The idea is that positions in a network could be defined on the basis of the similar intensity levels that the actors exhibit in expressing some features, instead of just considering relationships that actors hold with each others. Moreover, we propose a new clustering procedure that exploits the potentiality of copula functions, a mathematical instrument for the modelization of the stochastic dependence structure. Our clustering algorithm can be applied both to binary and real-valued matrices. We validate it with simulations and applications to real-world data. [ABSTRACT FROM AUTHOR]

In this paper we present an analysis of the behavior of Italian Twitter users during national political elections. We monitor the volumes of the tweets related to the leaders of the various political parties and we compare them to the elections results. Furthermore, we study the topics that are associated with the co-occurrence of two politicians in the same tweet. We cannot conclude, from a simple statistical analysis of tweet volume and their time evolution, that it is possible to precisely predict the election outcome (or at least not in our case of study that was characterized by a “too-close-to-call” scenario). On the other hand, we found that the volume of tweets and their change in time provide a very good proxy of the final results. We present this analysis both at a national level and at smaller levels, ranging from the regions composing the country to macro-areas (North, Center, South). [ABSTRACT FROM AUTHOR]

We investigate the impact of borders on the topology of spatially embedded networks. Indeed territorial subdivisions and geographical borders significantly hamper the geographical span of networks thus playing a key role in the formation of network communities. This is especially important in scientific and technological policy-making, highlighting the interplay between pressure for the internationalization to lead towards a global innovation system and the administrative borders imposed by the national and regional institutions. In this study we introduce an outreach index to quantify the impact of borders on the community structure and apply it to the case of the European and US patent co-inventors networks. We find that (a) the US connectivity decays as a power of distance, whereas we observe a faster exponential decay for Europe; (b) European network communities essentially correspond to nations and contiguous regions while US communities span multiple states across the whole country without any characteristic geographic scale. We confirm our findings by means of a set of simulations aimed at exploring the relationship between different patterns of cross-border community structures and the outreach index. [ABSTRACT FROM AUTHOR]

We analyze weighted networks as randomly reinforced urn processes, in which the edge-total weights are determined by a reinforcement mechanism. We develop a statistical test and a procedure based on it to study the evolution of networks over time, detecting the "dominance" of some edges with respect to the others and then assessing if a given instance of the network is taken at its steady state or not. Distance from the steady state can be considered as a measure of the relevance of the observed properties of the network. Our results are quite general, in the sense that they are not based on a particular probability distribution or functional form of the random weights. Moreover, the proposed tool can be applied also to dense networks, which have received little attention by the network community so far, since they are often problematic. We apply our procedure in the context of the International Trade Network, determining a core of "dominant edges." [ABSTRACT FROM AUTHOR]

The authors discuss the application of statistical physics to reconstruct business networks from limited credit data. They state that the precise structure of networks must be maintained to be adjustable to diffusion of innovations which are deeply entwined with its topology. The authors note that bankruptcy can result from propagated network distress.

The increasing attention to environmental issues is forcing the implementation of novel energy models based on renewable sources. This is fundamentally changing the configuration of energy management and is introducing new problems that are only partly understood. In particular, renewable energies introduce fluctuations which cause an increased request for conventional energy sources to balance energy requests at short notice. In order to develop an effective usage of low-carbon sources, such fluctuations must be understood and tamed. In this paper we present a microscopic model for the description and for the forecast of short time fluctuations related to renewable sources in order to estimate their effects on the electricity market. To account for the inter-dependencies in the energy market and the physical power dispatch network, we use a statistical mechanics approach to sample stochastic perturbations in the power system and an agent based approach for the prediction of the market players’ behavior. Our model is data-driven; it builds on one-day-ahead real market transactions in order to train agents’ behaviour and allows us to deduce the market share of different energy sources. We benchmarked our approach on the Italian market, finding a good accordance with real data. [ABSTRACT FROM AUTHOR]

Production systems, traditionally analyzed as almost independent national systems, are increasingly connected on a global scale. Only recently becoming available, the World Input-Output Database (WIOD) is one of the first efforts to construct the global multi-regional input-output (GMRIO) tables. By viewing the world input-output system as an interdependent network where the nodes are the individual industries in different economies and the edges are the monetary goods flows between industries, we analyze respectively the global, regional, and local network properties of the so-called world input-output network (WION) and document its evolution over time. At global level, we find that the industries are highly but asymmetrically connected, which implies that micro shocks can lead to macro fluctuations. At regional level, we find that the world production is still operated nationally or at most regionally as the communities detected are either individual economies or geographically well defined regions. Finally, at local level, for each industry we compare the network-based measures with the traditional methods of backward linkages. We find that the network-based measures such as PageRank centrality and community coreness measure can give valuable insights into identifying the key industries. [ABSTRACT FROM AUTHOR]

Theory of complex networks proved successful in the description of a variety of complex systems ranging from biology to computer science and to economics and finance. Here we use network models to describe the evolution of a particular economic system, namely the International Trade Network (ITN). Previous studies often assume that globalization and regionalization in international trade are contradictory to each other. We re-examine the relationship between globalization and regionalization by viewing the international trade system as an interdependent complex network. We use the modularity optimization method to detect communities and community cores in the ITN during the years 1995–2011. We find rich dynamics over time both inter- and intra-communities. In particular, the Asia-Oceania community disappeared and reemerged over time along with a switch in leadership from Japan to China. We provide a multilevel description of the evolution of the network where the global dynamics (i.e., communities disappear or reemerge) and the regional dynamics (i.e., community core changes between community members) are related. Moreover, simulation results show that the global dynamics can be generated by a simple dynamic-edge-weight mechanism. [ABSTRACT FROM AUTHOR]

Network analysis of functional imaging data reveals emergent features of the brain as a function of its topological properties. However, the brain is not a homogeneous network, and the dependence of functional connectivity parameters on neuroanatomical substrate and parcellation scale is a key issue. Moreover, the extent to which these topological properties depend on underlying neurochemical changes remains unclear. In the present study, we investigated both global statistical properties and the local, voxel-scale distribution of connectivity parameters of the rat brain. Different neurotransmitter systems were stimulated by pharmacological challenge (d-amphetamine, fluoxetine, and nicotine) to discriminate between stimulus-specific functional connectivity and more general features of the rat brain architecture. Although global connectivity parameters were similar, mapping of local connectivity parameters at high spatial resolution revealed strong neuroanatomical dependence of functional connectivity in the rat brain, with clear differentiation between the neocortex and older brain regions. Localized foci of high functional connectivity independent of drug challenge were found in the sensorimotor cortices, consistent with the high neuronal connectivity in these regions. Conversely, the topological properties and node roles in subcortical regions varied with neurochemical state and were dependent on the specific dynamics of the different functional processes elicited. [ABSTRACT FROM AUTHOR]

The aim of this paper is to study the dynamics of the commuting system of two insular regions of Italy, Sardinia and Sicily, inspected as complex networks. The authors refer to a 20-year time period and take into account three census data sets about the work and study-driven inter-municipal origin-destination movements of residential inhabitants in 1981, 1991 and 2001. Since it is likely that the number of municipalities (in this case, the vertices of the system) does not display sharp variations, the authors direct the study to the variation of the properties emerging through both a topological and a weighted network representation of commuting in the time periods indicated. [ABSTRACT FROM AUTHOR]

We study the structure of the network representing the interurban commuting traffic of the Sardinia region, Italy, which amounts to 375 municipalities and 1 600 000 inhabitants. We use a weighted network representation in which vertices correspond to towns and the edges correspond to the actual commuting flows among those towns. We characterize quantitatively both the topological and weighted properties of the resulting network. Interestingly, the statistical properties of the commuting traffic exhibit complex features and nontrivial relations with the underlying topology. We characterize quantitatively the traffic backbone among large cities and we give evidence for a very high heterogeneity of the commuter flows around large cities. We also discuss the interplay between the topological and dynamical properties of the network as well as their relation with sociodemographic variables such as population and monthly income. This analysis may be useful at various stages in environmental planning and provides analytical tools for a wide spectrum of applications ranging from impact evaluation to decision making and planning support. [ABSTRACT FROM AUTHOR]

This study investigated social visual attention in children with Autism Spectrum Disorder (ASD) and with typical development (TD) in the light of Brockmann and Geisel's model of visual attention. The probability distribution of gaze movements and clustering of gaze points, registered with eye-tracking technology, was studied during a free visual exploration of a gaze stimulus. A data-driven analysis of the distribution of eye movements was chosen to overcome any possible methodological problems related to the subjective expectations of the experimenters about the informative contents of the image in addition to a computational model to simulate group differences. Analysis of the eye-tracking data indicated that the scanpaths of children with TD and ASD were characterized by eye movements geometrically equivalent to Lévy flights. Children with ASD showed a higher frequency of long saccadic amplitudes compared with controls. A clustering analysis revealed a greater dispersion of eye movements for these children. Modeling of the results indicated higher values of the model parameter modulating the dispersion of eye movements for children with ASD. Together, the experimental results and the model point to a greater dispersion of gaze points in ASD [ABSTRACT FROM AUTHOR]

The role of Network Theory in the study of the financial crisis has been widely spotted in the latest years. It has been shown how the network topology and the dynamics running on top of it can trigger the outbreak of large systemic crisis. Following this methodological perspective we introduce here the Accounting Network, i.e. the network we can extract through vector similarities techniques from companies’ financial statements. We build the Accounting Network on a large database of worldwide banks in the period 2001–2013, covering the onset of the global financial crisis of mid-2007. After a careful data cleaning, we apply a quality check in the construction of the network, introducing a parameter (the Quality Ratio) capable of trading off the size of the sample (coverage) and the representativeness of the financial statements (accuracy). We compute several basic network statistics and check, with the Louvain community detection algorithm, for emerging communities of banks. Remarkably enough sensible regional aggregations show up with the Japanese and the US clusters dominating the community structure, although the presence of a geographically mixed community points to a gradual convergence of banks into similar supranational practices. Finally, a Principal Component Analysis procedure reveals the main economic components that influence communities’ heterogeneity. Even using the most basic vector similarity hypotheses on the composition of the financial statements, the signature of the financial crisis clearly arises across the years around 2008. We finally discuss how the Accounting Networks can be improved to reflect the best practices in the financial statement analysis. [ABSTRACT FROM AUTHOR]

The fragmentation of production across countries has become an important feature of the globalization in recent decades and is often conceptualized by the term “global value chains” (GVCs). When empirically investigating the GVCs, previous studies are mainly interested in knowing how global the GVCs are rather than how the GVCs look like. From a complex networks perspective, we use the World Input-Output Database (WIOD) to study the evolution of the global production system. We find that the industry-level GVCs are indeed not chain-like but are better characterized by the tree topology. Hence, we compute the global value trees (GVTs) for all the industries available in the WIOD. Moreover, we compute an industry importance measure based on the GVTs and compare it with other network centrality measures. Finally, we discuss some future applications of the GVTs. [ABSTRACT FROM AUTHOR]

We analyze the network of relations between parliament members according to their voting behavior. In particular, we examine the emergent community structure with respect to political coalitions and government alliances. We rely on tools developed in the Complex Network literature to explore the core of these communities and use their topological features to develop new metrics for party polarization, internal coalition cohesiveness and government strength. As a case study, we focus on the Chamber of Deputies of the Italian Parliament, for which we are able to characterize the heterogeneity of the ruling coalition as well as parties specific contributions to the stability of the government over time. We find sharp contrast in the political debate which surprisingly does not imply a relevant structure based on established parties. We take a closer look to changes in the community structure after parties split up and their effect on the position of single deputies within communities. Finally, we introduce a way to track the stability of the government coalition over time that is able to discern the contribution of each member along with the impact of its possible defection. While our case study relies on the Italian parliament, whose relevance has come into the international spotlight in the present economic downturn, the methods developed here are entirely general and can therefore be applied to a multitude of other scenarios. [ABSTRACT FROM AUTHOR]

This work analyzes methods for the identification and the stability under perturbation of a territorial community structure with specific reference to transportation networks. We considered networks of commuters for a city and an insular region. In both cases, we have studied the distribution of commuters' trips (i.e., home-to-work trips and vice versa). The identification and stability of the communities' cores are linked to the land-use distribution within the zone system, and therefore their proper definition may be useful to transport planners. [ABSTRACT FROM AUTHOR]

Community detection is an important tool for exploring and classifying the properties of large complex networks and should be of great help for spatial networks. Indeed, in addition to their location, nodes in spatial networks can have attributes such as the language for individuals, or any other socio-economical feature that we would like to identify in communities. We discuss in this paper a crucial aspect which was not considered in previous studies which is the possible existence of correlations between space and attributes. Introducing a simple toy model in which both space and node attributes are considered, we discuss the effect of space-attribute correlations on the results of various community detection methods proposed for spatial networks in this paper and in previous studies. When space is irrelevant, our model is equivalent to the stochastic block model which has been shown to display a detectability-non detectability transition. In the regime where space dominates the link formation process, most methods can fail to recover the communities, an effect which is particularly marked when space-attributes correlations are strong. In this latter case, community detection methods which remove the spatial component of the network can miss a large part of the community structure and can lead to incorrect results. [ABSTRACT FROM AUTHOR]