Search

Being able to learn from complex data with phase information is imperative for many signal processing applications. Today’s real-valued deep neural networks (DNNs) have shown efficiency in latent information analysis but fall short when applied to the complex domain. Deep complex networks (DCN), in contrast, can learn from complex data, but have high computational costs; therefore, they cannot satisfy the instant decision-making requirements of many deployable systems dealing with short observations or short signal bursts. Recent, Binarized Complex Neural Network (BCNN), which integrates DCNs with binarized neural networks (BNN), shows great potential in classifying complex data in real-time. In this paper, we propose a structural pruning based accelerator of BCNN, which is able to provide more than 5000 frames/s inference throughput on edge devices. The high performance comes from both the algorithm and hardware sides. On the algorithm side, we conduct structural pruning to the original BCNN models and obtain 20 × pruning rates with negligible accuracy loss; on the hardware side, we propose a novel 2D convolution operation accelerator for the binary complex neural network. Experimental results show that the proposed design works with over 90% utilization and is able to achieve the inference throughput of 5882 frames/s and 4938 frames/s for complex NIN-Net and ResNet-18 using CIFAR-10 dataset and Alveo U280 Board.

The exponential increase in the number of scientific publications raises the question of whether the sciences are expanding into a fractured structure, making cross-field communication difficult. On the other hand, scientists may be motivated to learn extensively across fields to enhance their innovative capacity, and this may offset the negative effects of fragmentation. Through an investigation of the distances within and clustering of cross-sectional citation networks, this study presents evidence that fields of science become more integrated over time. The average citation distance between papers published in the same year decreased from approximately 5.33 to 3.18 steps between 1950 and 2018. This observation is attributed to the growth of cross-field communication throughout the entire period as well as the growing importance of high impact papers to bridge networks in the same year. Three empirical findings support this conclusion. First, distances decreased between almost all disciplines throughout the time period. Second, inequality in the number of citations received by papers increased, and as a consequence the shortest paths in the network depend more on high impact papers later in the period. Third, the dispersion of connections between fields increased continually. Moreover, these changes did not entail a lower level of clustering of citations. Both within- and cross-field citations show a similar rate of slowly growing clustering values in all years. The latter findings suggest that domain spanning scholarly communication is partly enabled by new fields that connect disciplines.

Recent work has employed information theory in social and complex networks. Studies often discuss entropy in the degree distributions of a network. However, no specific work on entropy exists in clique networks. This work is an extension of a previous study that discussed this topic. We propose a method for calculating the entropy of a clique network and its minimum and maximum values in temporal semantic networks based on titles of scientific papers. In addition, the critical network of moments was extracted. We use the titles of scientific papers published in Nature and Science over ten-year period. The results show the diversity of vocabulary over time, based on the entropy values of vertices and edges. In each critical network, we discover the paths that connect important words and an interesting modular structure.

As an important recyclable and reusable resource, waste paper is traded in millions of dollars around the world every year. Global waste paper trade not only addresses resource scarcity issues, alleviates environmental pressures and brings substantial economic gains, but also contributes to the development of global circular economy. Using complex network methods, bilateral waste paper trade data, and temporal exponential random graph models (TERGM), we construct global waste paper trade networks (GWPTNs) during 2000–2018, and examine their structural evolution and determinants. We report that: (I) GWPTNs display obvious features of small-world network, low reciprocity, heterogeneity and disassortativity; (II) Asia is the leading recipient of global waste paper, while Europe and North America are the main exporters; (III) China and India dominate the waste paper import markets while the United States is the largest exporter, and Germany plays an essential role in both importing and exporting hubs of waste paper; (IV) The evolution of the GWPTNs is significantly influenced by endogenous reciprocity, transitivity and preferential attachment, and economies with more partners or within the same continents are more likely to trade waste paper. Economies with higher urbanization rates, more per capita income, stricter environmental regulation, or lower industrialization rates are more likely to export waste paper; conversely, they are more likely to be the importers. Economy-pairs sharing a language or religion, being a former colony of the same colonizer, historical colonial relationship or a border, or signing regional trade agreements are more likely to trade waste paper.

Background To sustain the critical progress made, prioritization and a multidisciplinary approach to malaria research remain important to the national malaria control program in Benin. To document the structure of the malaria collaborative research in Benin, we analyze authorship of the scientific documents published on malaria from Benin. Methods We collected bibliographic data from the Web Of Science on malaria research in Benin from January 1996 to December 2016. From the collected data, a mulitigraph co-authorship network with authors representing vertices was generated. An edge was drawn between two authors when they co-author a paper. We computed vertex degree, betweenness, closeness, and eigenvectors among others to identify prolific authors. We further assess the weak points and how information flow in the network. Finally, we perform a hierarchical clustering analysis, and Monte-Carlo simulations. Results Overall, 427 publications were included in this study. The generated network contained 1792 authors and 116,388 parallel edges which converted in a weighted graph of 1792 vertices and 95,787 edges. Our results suggested that prolific authors with higher degrees tend to collaborate more. The hierarchical clustering revealed 23 clusters, seven of which form a giant component containing 94% of all the vertices in the network. This giant component has all the characteristics of a small-world network with a small shortest path distance between pairs of three, a diameter of 10 and a high clustering coefficient of 0.964. However, Monte-Carlo simulations suggested our observed network is an unusual type of small-world network. Sixteen vertices were identified as weak articulation points within the network. Conclusion The malaria research collaboration network in Benin is a complex network that seems to display the characteristics of a small-world network. This research reveals the presence of closed research groups where collaborative research likely happens only between members. Interdisciplinary collaboration tends to occur at higher levels between prolific researchers. Continuously supporting, stabilizing the identified key brokers and most productive authors in the Malaria research collaborative network is an urgent need in Benin. It will foster the malaria research network and ensure the promotion of junior scientists in the field.

We consider scientific publications as a growing complex network where papers are nodes and citations are links which connect these papers together. We explain this network approach and make special accent on the temporal aspect of citation network, namely, we focus on the growth of the number of papers, age distribution of references, and citation dynamics. We trace relation of the age distribution of references to citation dynamics and explore this reference-citation duality.

The model of citation dynamics developed in Chap. 3 is compared to measurements in order to determine empirical functions and parameters such as aging function, obsolescence function, and the paper’s fitness. We found that the aging function is universal, namely, it is the same for all papers in one field published in the same year. However, the obsolescence function depends on the number of previous citations. This unexpected finding prompted us to focus more closely on the network aspect of citation dynamics and to consider not only the nearest neighbors of each paper in the citation network, but its next-nearest neighbors as well. The updated model takes into account the correlations between citation dynamics of a paper and its neighbors (network assortativity).

This paper builds on recent insights from network theory and on the rich dataset made available by the Panama Papers in order to investigate the micro-economic dynamics of tax-evasion. We model offshore financial entities documented in the Panama Papers as links between jurisdictions in the global network of tax evasion. A quantitative analysis shows that the resulting network, far from being a random collection of bilateral links, has key features of complex networks such as a core-periphery structure and a fat-tail degree distribution. We argue that these structural features imply that policy must adopt a systemic perspective on the matter. We offer three sets of insights from this perspective. First, we identify through centrality measures tax havens that ought to be priority policy targets. Second, we show that efficient tax treaties must contain exchange information clauses and link tax-havens to non-haven jurisdictions. Third, we show that the optimal deterrence strategies for a social-planner facing a strategic tax-evader in a Stackelberg competition can be characterized using the notion of Bonacich centrality.

This is a call for papers that should contribute to the unification of behavioral sciences and team management, focusing on the biological origin of cooperation and swarm intelligence, moving from biology to psychology and from sociology to political science, with the help of the theoretical tools of complex networks. This issue should shed light into the origin of ergodicity breaking and contribute to establishing a connection, still lacking theoretical support, between complexity properties that are expected to be correlated. Examples are: non-Poisson renewal events and multi-fractality; complexity matching and chaos synchronization; criticality and extended criticality of small size systems. Although the emphasis is on systems of small size, and especially on the search of the size maximizing both information transport and cooperation emergence, special attention will be devoted to the interaction between small groups and their larger societies.

The Panama Papers represent a large set of relationships between people, companies, and organizations that had affairs with the Panamanian offshore law firm Mossack Fonseca, often due to money laundering. In this paper, we address for the first time the problem of searching the Panama Papers for people and companies that may be involved in illegal acts. We use a collection of international blacklists of sanctioned people and organizations as ground truth for bad entities. We propose a new ranking algorithm, named Suspiciousness Rank Back and Forth (SRBF), that leverages this ground truth to assign a degree of suspiciousness to each entity in the Panama Papers. We experimentally show that our algorithm achieves an AUROC of 0.85 and an Area Under the Recall Curve of 0.87 and outperforms existing techniques.

The problem of community detection (or clustering) in graphs plays an important role in analysis of complex large-scale networks and big data structures, arising in natural, behavioral and engineering sciences. Examples of such networks include, but are not limited to, World Wide Web (WWW) and Internet, social networks, ecological networks and food webs, cellular and molecular ensembles. A community (or a module) in a graph is a subset of its nodes, whose members are "densely" connected to each other yet have relatively few connections with nodes outside this subset. A number of algorithms to subdivide the nodes of large-scale graphs into communities have recently been proposed; many of them hunt for the graph’s partitions of maximal modularity. One of the most efficient graph clustering algorithms of this type is the Multi-Level Aggregation (or "Louvain") method. In this paper, a randomized counterpart of this algorithm is proposed, which provides a comparable "quality" of graph’s clustering, being however much faster on huge graphs. We demonstrate the efficiency of our algorithm, comparing its performance on several "benchmark" large-scale graphs with existing methods.

Convolutional Neural Networks (CNNs) have been widely investigated as some of the most promising solution for various computer vision tasks. However, CNNs introduce massive computing overhead due to their complex network computing flow, resulting in significantly reduced applicability and performance, especially in the mobile devices. Various optimization schemes have been proposed mainly based on both model compression and stacked external computing resources. While these schemes have been proven effective, methods which take into account mobile-specific context-aware optimization approaches have been largely overlooked. One such opportunity is the feasible CNN computing flow simplification to the under-test objects with distinguish features, which can be efficiently pre-analyzed inside the mobile sensor system. Hence, we propose VoCaM, a visualization oriented CNN acceleration framework on mobile devices for image classification tasks. VoCaM takes advantage of the mobile camera system, where the comprehensive pre-analysis can be conducted to reveal the color composition of the under-test images without incurring any additional overhead. Also, the visualization analysis of VoCaM reveals that, certain color-specific filters may have very trivial result impact when the under-test images have mismatching primary color components. Then a set of approximate computing methods is applied to these insignificant filters to replace the intensive convolutional operation, and greatly accelerate the computing process. With ignorable overhead, VoCaM can significantly optimize the computation load of the convolutional layers, with very small impact on the overall classification accuracy.

Research on mutual cooperation among scholars or research institutions has become more and more common. Thepurpose of this paper is to explore the current status of cooperation between scholars and research institutions in thefield of Chinese education. In this paper, we use the method of the complex network to analyze the cooperativebehavior of academic papers published by Chinese educational scholars by collecting academic papers on educationleadership, education policy, quality education, and vocational education. Our conclusions show that most of theacademic papers published by Chinese educational scholars are non-cooperative. In the authors of the co-authoredpapers, there is a significant "Matthew effect", that is, some key scholars in these fields that link the collaborators.Lastly, there is no obvious aggregation effect between the authors of the co-authored papers which indicating awidespread and extensive connection between the collaborators. The above conclusions provide valuable insightsinto our understanding of the cooperative behavior of Chinese education scholars.

This paper proposes h l -index as an improvement of the h-index, a popular measurement for the research quality of academic researchers. Although the h-index integrates the number of publications and the academic impact of each publication to evaluate the productivity of a researcher, it assumes that all papers that cite an academic article contribute equally to the academic impact of this article. This assumption, of course, could not be true in most times. The citation from a well-cited paper certainly brings more attention to the article than the citation from a paper that people do not pay attention to. It therefore becomes important to integrate the impact of papers that cite a researcher's work into the evaluation of the productivity of the researcher. Constructing a citation network among academic papers, this paper therefore proposes h l -index that integrating the h-index with the concept of lobby index, a measures that has been used to evaluate the impact of a node in a complex network based on the impact of other nodes that the focal node has direct link with. This paper also explores the characteristics of the proposed h l -index by comparing it with citations, h-index and its variant g-index.

Complex robotic systems like the DLR Hand Arm System integrates a huge amount of sensors and actuators. Hence system design and especially communication infrastructure design has to be flexible in a heterogeneous network of different bus systems. As basis, a modular electronic concept as well as a well-structured communication concept is necessary [1]-[3]. SpaceWire suites well to these requirements since on one hand it supports arbitrary topologies from point to point up complex network structures and on the other hand it is easy to implement and has a small footprint. Additionally its logical and regional addressing scheme enables changes in the topology during runtime simply by reprogramming the routing switches. However, such changes require expert knowledge. This work presents a graphical method to setup and configure SpaceWire network topologies. This enables non-experts to replace or integrate new components to the system or to set up a test bed to investigate a specific aspect. The developer provides a GraphML description [4] specifying the SpaceWire communication capabilities of each component. Thus the user is able to adapt the SpaceWire network topology or to set up a new one simply by merging the different GraphML descriptions of the used components. A post process is afterwards used to analyze the GraphML description and to generate the necessary configuration messages according to the topology. This enables faster development cycles and rapid prototyping. The approach is approved and explained using the SpaceWire network topology of the DLR Hand Arm System.

In this paper we study the topological structure of semantic networks based on titles of papers published in scientific journals. It discusses its properties and presents some reflections on how the use of social and complex network models can contribute to the diffusion of knowledge. The proposed method presented here is applied to scientific journals where the titles of papers are in English or in Portuguese. We show that the topology of studied semantic networks are small-world and scale-free.