Earticle

외교의 행위주체들 간의 상호관계 양식은 국제관계의 역사적 전개과정에서 다양한 변화양상을 제기해 왔다. 전근대적 및 근대적 외교를 거쳐 현재의 탈근대화, 세계화, 지식정보화, 민주화 추세에 따라 외교 또한 지방화 시대에 직면하여 새로운 관념과 실행양식을 요구하고 있다. 본 연구는 9·11 테러를 계기로 중앙정부 차원에서 시도된 외교변환 모델에서 제기된 미국형 새로움을 기반으로 향후 지방외교 또는 도시외교의 방향 및 인식 등에 제공될 수 있는 발전적 함의를 탐색하고자 한다. ‘4개년 외교·개발검토보고서(QDDR)’를 중심으로 한 미국형 외교변환 모델을 통해 국제외교 환경의 새로운 지형, 외교의 주체와 상호관계, 쟁점영역의 연계성 및 실행방식 등을 중심으로 제기된 스마트파워형 복합네크워크 외교양식을 통해, 상대적으로 전통적인 유형 및 수준에 머물러 있는 현 단계 지방외교의 미래형과 연계될 수 있는 함의 제시가 유의미한 기대로 발전될 수 있음을 모색하고자 한다. 다만 현재의 지방외교에서 노출된 탈근대형 외교에 대한 인식 및 관심의 부족, 전문인력·프로그램·예산 등 실행 체계 및 효율적 운용 능력의 부재 등 지금까지 일관되게 관찰되어 온 한계점들을 전제한 범위 내에서 논의될 것이다.

The Interaction modes among general diplomatic agents has been presented in various forms during historical development of international relations. After pre-modern and modern diplomacy modes, the up-to-date diplomatic environments, including globalization and knowledge-based information, requires new creative ideas and practices of diplomacy. This study aims to analyzes the ever-changing diplomatic configuration and provides some evolutionary implications toward local diplomacy, through some newness of the U.S. transformational diplomacy on its central government's base since September 11 terrorist reaction. Those implications includes, through analyzing the U.S.'s Transformational Diplomacy and QDDR, some emerging configuration of international diplomatic environment, interrelationship among diplomatic agents, reciprocal linkages among diplomatic issue-areas and their practices. In spite of some significant implications, however, found toward future orientations for local diplomacy, this study is based on the recognition that present local diplomacy has some deep-rooted weakness and limitations in terms of adequate perception of changing diplomacy, inexistence of specialized expertize, vulnerable budget system, inefficient operational capacity of relevant programs, etc.

웹에 존재하는 악성코드 배포 네트워크에는 악성코드 배포를 위해 핵심 역할을 수행하는 중심 노드가 있다. 이 노드를 찾아 차단하면 악성코드 전파를 효과적으로 차단할 수 있다. 본 연구에서는 복잡계 네트워크에서 위험 분 석이 적용된 centrality 검색 방법을 제안하였고, 이 방식을 통해 악성코드 배포 네트워크 내에서 핵심노드를 찾는 방법을 소개한다. 그 외에, 정상 네트워크와 악성 네트워트는 in-degree와 out-degree 측면에서 큰 차이가 있고, 네트워크 레이아웃 측면에서도 서로 다르다. 이 특징을 통해 우리는 악성과 정상 네트워크를 분별할 수 있다.

In the malware distribution network existing on the web, there is a central node that plays a key role in distributing malware. If you find and block this node, you can effectively block the propagation of malware. In this study, a centrality search method applied with risk analysis in a complex network is proposed, and a method for finding a core node in a malware distribution network is introduced through this approach. In addition, there is a big difference between a benign network and a malicious network in terms of in-degree and out-degree, and also in terms of network layout. Through these characteristics, we can discriminate between malicious and benign networks.

인류는 다양한 복잡계(Complex Networks) 속에서 살아가고 있어서 이를 잘 해석하는 것은 자연계에서의 대두되는 문제를 해결하는 실마리를 찿을 수 있는 길이라고 본다. 트위트, 페이스북 등과 같은 소셜네트워크망의 구성도, 항공기 등 교통수단 구성도, 각종 통신선로망, 유전자의 대사작용, 질병의 발병/전파도, 비즈니스 모델 구성 등이 복잡계(Complex Networks)의 사례들이다. 인문사회 또는 자연과학 분야, 공학 분야 등 전분야에서 복잡계를 쉽게 분석 또는 설명하고자 많은 이론과 모델(ER Random network 모델, SIR 모델, Watts-Strogatz 모델, Chung-Lu 모델, Preferential Attachment 모델 등)이 개발되어 사용되어 지고 있으나 복잡계를 완전하게 표시하지는 못하고 있다. 이를 직관적으로 이해할 수 있도록 PCB(Propagation Communication Board; 정보 전파 보드) 모델을 제안하였다. 이 제안 모델을 검증하기 위해서 복잡계에 적용 가능한 사례를 도입하여 검증을 해 보았다. 우선적으로 교통/운송망에 적용하여 정보 전달이 어떻게 이루어지는지 살펴보고, 유무선 통신망에서의 정보 전달 과정을, 마지막으로 홍보 매체에서의 홍보용 정보가 어떻게 전달되는지를 보았다. 본 논문에서는 언급하지 않은 질병 또는 전염병의 전달 경로 등 다른 복잡계에서도 본 제안 모델이 적용 가능하다.

This study combines the concept of degree centrality in complex network with the Term Frequency $^*$ Proportional Document Frequency ($TF^*PDF$) algorithm; the combined method, called PMCN (PDF-Modified similarity and Complex Network), constructs relationship networks among sentences for writing news summaries. The PMCN method is a multi-document summarization extension of the ideas of Bun and Ishizuka (2002), who first published the $TF^*PDF$ algorithm for detecting hot topics. In their $TF^*PDF$ algorithm, Bun and Ishizuka defined the publisher of a news item as its channel. If the PDF weight of a term is higher than the weights of other terms, then the term is hotter than the other terms. However, this study attempts to develop summaries for news items. Because the $TF^*PDF$ algorithm summarizes daily news, PMCN replaces the concept of "channel" with "the date of the news event", and uses the resulting chronicle ordering for a multi-document summarization algorithm, of which the F-measure scores were 0.042 and 0.051 higher than LexRank for the famous d30001t and d30003t tasks, respectively.

In modern economy, any firm cannot sustain by itself. Whatever the type of a firm, it is tangled up with other firms in any relation of supply and demand, partnership or competition, etc. In 2021, for example, chip shortage caused by the spread of COVID-19 and the sanction on Chinese chipmakers stroke the global automotive industry. As a result, major car original equipment manufacturers (OEMs) such as GM, Ford, Toyota, and many others stopped their production, which serially stopped their automotive component supplier’s production as well. However, the business magazine Forbes warns the worse situation with the messages “Battery scarcity could dwarf chip shortage impact on global auto sales” and “The semiconductor shortage will cut a total of 8.1 million cars from global production between 2021 and 2023, while between 2022 and 2029, 18.7 million rechargeable electric cars will be lost because of battery cell shortages” (Winton 2021). Due to the interdependence between firms, if any disruption erupts in the industry, its ripple effect impacts the cross-over in a multi-tiered value chain (Cachon and Lariviere, 2001). In 2010s, electric vehicle industry emerged by strong environment regulations and the hyper growth of the industry pioneers who are strongly based on information technology (IT), such as Tesla. Actually, automotive industry including electric vehicle is no longer limited to manufacturing, but is growing rapidly through a complex value chain combined with advanced IT technologies. Therefore it is highly required to understand the complex pattern and sustainability of the highly entangled system. In this research we investigate how the electric vehicle industry emerges and the resilience of the value chain changes. To tackle the research question, we introduce the novel methodological framework based on complex network theory and agent based model approach. From the empirical data on electric vehicle value chain of 6 years, we apply the framework to address the evolution of the industry using weighted network analysis (Figure 1). We found that the electric vehicle industry becomes more heterogeneous structure (Figure 2). This means that the companies. In complex network theory, the heterogeneous network structure has been well known to have the properties of “double-edge swords” in the robustness of the system, that is, it is usually very robust, but the collapse of a few central firms can cause whole system to collapse with a huge ripple effect (Albert et al. 2000). With the understandings of structural properties of electric vehicle value chain network, we examine the resilience of the network with introducing the agent based simulation model. From our simulation results, as year goes, the electric vehicle value chain becomes less resilient. Furthermore, from the result of the distinction between individual firm’s impact of unweighted and weighted aspect, we found some firm’s influence would be underestimated when one only consider the unweighted network properties. Our novel approach based on weight network analysis and agent based modeling would more capture the characteristics of electric vehicle industry and shed light on understanding systemic risks due to the individual firm’s disruption.

지속적인 기후 및 도시구조의 변화, 기술의 발전 등으로 인해 최근의 재난들은 종래의 전통적 재난 유형과 다른 양상으로 발현되고 있으며, 사회 구성요소 간 연결성의 증대로 재난은 더욱 대형화․복잡 화된 특징을 띄고 있다. 본 연구에서는 복합재난의 핵심 개념인 재난의 “연계성”을 파악하기 위해, 전문가 조사를 통해 구축한 재난유형별 취약요소 정보를 기반으로 사회연결망 분석을 실시하였다. 네트워크 중심성 분석결과, 풍수해, 미세먼지 등은 모든 종류의 중심성이 높아 재난의 연계성 관점에 서 매우 중요한 재난유형이라 할 수 있다. 연결 중심성이 높은 감염병, 정보통신은 사회 전반적인 유기적인 대책이 요구되며, 근접 중심성이 높은 사회기반시설 붕괴, 산불의 경우 타 재난으로 연계되 는 경로가 짧아 신속한 대응이 필요한 것으로 나타났다. 커뮤니티 분석을 통해 재난의 연계성 관점 에서 전체 재난유형이 6개의 그룹으로 분류되는 것을 확인하였고, 이는 발생 가능성이 상대적으로 높은 복합재난 상황을 가정할 수 있는 기초자료로 활용될 수 있을 것으로 판단된다.

Recent disasters are emerging in different ways from traditional disaster types due to climate change, urban structural changes, and technological advances. Disasters are becoming larger and more complex because of the increased connectivity between social components. In this study, social network analysis was conducted to identify the “connectivity” of disasters based on information of vulnerabilities by type of disaster established through expert surveys. As a result of the analysis of network centrality, storm and flood damage, fine particulate matter etc. are significant types of disasters in terms of disaster linkages since they were highly calculated for all kinds of centrality. Infectious diseases and information and communication accidents with a high degree centrality index require systematic measures across society. Infrastructure collapse and forest fires with high closeness centrality require a prompt response since the path to other disasters is relatively short. Through community analysis, it was confirmed that the all disaster types were classified into six groups in terms of the connection of disasters. This can be used as basic data to assume complex disaster situations with relatively high probability of occurrence.

최근 높은 정확도를 보이는 복잡한 신경망 모델들이 직면한 주요 과제는 다중 GPU 시스템에서의 효율적인 배포입니다. 신경망 모델의 복잡한 계층 간 의존성과 다중 GPU 시스템의 가변적인 데이터 통신 오버헤드는 수동 스케줄링 방식으로는 일관된 성능 향상을 이루기 어렵게 만듭니다. 이 문제를 해결하기 위해, 본 논문에서는 NN Maestro라는 새로운 계층 스케줄링 접근 방식을 제안합니다. 이 방식은 다중 GPU 시스템에서 데이터 통신 오버헤드를 최소화함으로써, 복잡한 신경망 모델의 추론 지연 시간을 개선하는 효율적인 병렬 실행 전략을 제공합니다. NN Maestro는 사전 훈련된 SVM 분류기를 통해 다중 GPU 스케줄링의 이점을 평가하고, 계층의 스케줄링 순서를 Topological Sort와 Significance Cost를 바탕으로 계산합니다. 이후 NN Maestro는 Placement Cost를 비교해 최적의 GPU를 선택하고, 병렬 실행을 위해 계층을 Grouping하여 최종 스케줄링 결과를 생성합니다. 다양한 다중 GPU 구성 (2080Ti 2개, V100 4개, A6000 4개 GPU)에서, NN Maestro는 기준 성능 대비 최대 1.67배의 성능 향상을 달성합니다.

The main challenge facing recent complex neural network models, which have shown competitive accuracy, is their efficient deployment in multi-GPU systems. The complex inter-layer dependences of the neural network models combined with the variable data communication overhead of multi-GPU systems make it almost impossible to achieve a fair performance gain under manual scheduling. To address this problem, we propose a new layer-scheduling approach called NN Maestro, which generates an efficient parallel execution strategy for multi-GPU systems that minimizes the data communication overhead, thereby improving the inference latency for complex neural network models. NN Maestro evaluates the advantages of multi-GPU scheduling using a pre-trained SVM classifier and calculates the scheduling order of layers based on Topological Sort and Significance Cost. Then, NN Maestro selects the optimal GPU by comparing Placement Costs and generates the final scheduling result by grouping layers for parallel execution. On various multi-GPU configurations (2 2080Ti, 4 V100, and 4 A6000 GPUs), NN Maestro achieves up to 1.67x of performance improvement over the baseline.

The purpose of this study was to analyze the impact of greenhouse cultivation area and groundwater level changes due to the water curtain cultivation in the greenhouse complexes, which are mainly situated along rivers where water resources are easy to secure. The groundwater observation network in Miryang, Gyeongsangnam-do, located downstream of the Nakdong River, was selected for the study area. We classified the groundwater monitoring well into the greenhouse (riverside) and field cultivation areas (plain and mountain) to compare the groundwater impact of water curtain cultivation in the greenhouse complex. The characteristics of groundwater level changes classified by terrain type were analyzed using the observed data. Riverside wells have significant permeability coefficients and are close to rivers, so they are greatly affected by river flow and precipitation changes so that water level shows a specific pattern of annual changes. Most plain wells do not show a constant annual change, but observation wells near small rivers and small-scale greenhouse cultivation areas sometimes show annual and daily changes in which the water level drops during winter. Compared to other observation wells, mountain wells do not show significant yearly changes in water level and show general characteristics of bedrock aquifer well with a low permeability coefficient.

In this paper, we present a new method to tool path planning problem for rough cut of pocket milling operations. The key idea is to formulate the tool path problem into a TSP (Travelling Salesman Problem) so that the powerful neural network approach can be effectively applied. Specifically, our method is composed of three procedures: a) discretization of the pocket area into a finite number of tool points, b) neural network approach (called SOM-Self Organizing Map) for path finding, and c) postprocessing for path smoothing and feedrate adjustment. By the neural network procedure, an efficient tool path (in the sense of path length and tool retraction) can be robustly obtained for any arbitrary shaped pockets with many islands. In the postprocessing, a) the detailed shape of the path is fine tuned by eliminating sharp corners of the path segments, and b) any cross-overs between the path segments and islands. With the determined tool path, the feedrate adjustment is finally performed for legitimate motion without requiring excessive cutting forces. The validity and powerfulness of the algorithm is demonstrated through various computer simulations and real machining.

연구목적: 도시의 물리적·비물리적 구조간 연결이 확대되고 복잡해짐에 따라 재난으로 인한 피해가 복 합적으로 발생하는 복합재난의 위험성이 증가하고 있다. 이러한 복합재난에 대비하기 위해서는 복합 재난으로 전개될 수 있는 재난들을 선제적으로 식별하여 관리하는 것이 중요하다. 이러한 배경에서 본 연구는 국내외 복합재난 관련 연구의 동향을 분석함으로써 복합재난으로 연구된 재난 유형을 분석하 고, 이를 통해 향후 복합재난 관리의 방향성을 제시하고자 한다. 연구방법: 본 연구는 재난과 관련된 국 내외 학술지에 최근 20년(2002-2021년)간 등재된 복합재난 관련 993편의 서지정보에 기반하여 동시출 현 빈도분석을 수행하여 재난 유형 간 네트워크를 구축하였으며, 네트워크 분석을 통해 복합재난 연구 동향에 관한 국내외 및 시기별 비교분석을 수행하였다. 연구결과: 국내에서는 풍수해(집중호우, 태풍등), 기반시설 붕괴, 화재와 관련한 복합재난 연구의 비중이 높았으며, 최근 들어 지진과 산사태와 관련된 복합재난 연구가 증가하는 것으로 분석되었다. 반면, 국외에서는 풍수해 및 지진과 더불어 기반시설 붕괴에 관한 연구의 비중이 높았으며, 지진해일과 정전 등 재난 연계 유형 이 다양하게 나타났다. 결론: 본 연구는 복합재난 연구 동향에 대한 이해도를 높이고, 앞으로 국내 복합재난 연구가 가져야 할 방향성을 제안 하는 데 활용할 수 있을 것으로 기대된다.

Purpose: As the connection between physical and non-physical structures in cities is expanding and becoming more complex, the risk of complex disaster which causes damage in a complex way is increasing. Preparing for these complex disasters, it is important to preemptively identify and manage disasters that can develop into complex disasters. Therefore, this study analyzes the disaster types studied as complex disasters by analyzing the trends of domestic and international studies related to complex disasters, and presents the direction of complex disaster management in the future. Method: We first established co-occurrence networks between disaster types based on 993 articles related to complex disasters published in disaster-related journals for the last 20 years (2002-2021). Then, through network analysis, domestic and international complex disaster research trends were compared and analyzed. Result: Research on complex disasters related to storm and flood damage, infrastructure failure and fire was high in domestic studies, and it was analyzed that research on complex disasters related to earthquakes and landslides has recently increased. However, in international studies, the proportion of studies on infrastructure failure along with storm and flood damage and earthquake was high, and various types of disasters such as tsunami and drought appeared. Conclusion: The results of this study are expected to increase the understanding of the trends in complex disaster research and provide suggestions of domestic complex disaster research in the future.

본 연구는 인천 남동국가산업단지 입주기업을 대상으로 판매 네트워크와 기업 간 판매거래량의 결정요인을 분석하였다. NICE신용평가정보의 부가가치세 거래자료(2019–2021)를 활용해 판매 연계의 공간·산업 분포를 파악하고, 매출 상위 100개 선도기업 중심의 다층 공급망을 재구성해 남동산단 기업의 단계별 위상을 확인하며, 중력모형을 추정하여 판매거래량 결정요인을 식별하였다. 주요 결과는 다음과 같다. 첫째, 남동산단의 판매 네트워크는 수도권에 집중된 구조이나 산업단지 내부 거래는 4.02%에 불과하다. 업종 측면에서는 전자부품·기계장비가 핵심 축을 이루고, 이들 업종은 동일 업종 내 거래비중이 높을 뿐 아니라 전기장비·자동차·금속가공 등 연관 업종과의 거래가 활발하다. 둘째, 선도기업 중심의 공급망에서 남동산단 기업은 주로 2차와 3차 단계에서 참여하며, 공급망의 하위 단계로 내려갈수록 평균 공급 비중이 증가한다. 특히 3차 공급망에서는 일부 조합에서 공급 비중이 100%에 달하는 등 특정 거래에서 독점적 지위가 확인된다. 셋째, 판매거래량 결정요인은 대체로 중력모형의 가정을 따르나, 선도기업 중심 공급망 내부에서는 지리적 근접성의 효과가 유의하지 않고, 대신 거래집중도가 판매규모 확대에 유의한 양(+)의 영향을 보인다. 이는 공급망 맥락에서 관계의 안정성과 지속성이 거래강도를 높일 수 있음을 의미한다. 이러한 결과는 수도권 연계 물류·정보 인프라의 광역적 고도화, 연관 업종 간 협력 강화, 장기·안정적(관계특화) 거래 촉진 등 정책 설계에 시사점을 제공한다. 이러한 결과는 수도권 연계 물류·정보 인프라의 광역적 고도화, 연관 업종 간 협력 강화, 장기·안정적(관계특화) 거래 촉진 등 산업단지의 기능적 연계구조 강화를 위한 정책 설계에 시사점을 제공한다.

This study examines the sales network and determinants of inter-firm sales volume among firms in the Incheon Namdong National Industrial Complex. Using NICE Credit Information’s value-added tax transaction data (2019-2021), we map the spatial and industrial distribution of sales ties, reconstruct a lead-firm-centered supply chain to locate Namdong firms within it, and estimate a gravity model to identify determinants of sales volume. Key findings are as follows. First, internal transactions account for only 4.02%, with most sales flowing to the Seoul–Gyeonggi area. Second, electronics components and machinery form the core industrial axes linking to related sectors. Third, in the lead-firm chain, Namdong firms participate mainly as 2nd- and 3rd-tier suppliers, with average supply shares increasing at lower tiers. Fourth, within lead-firm transactions, seller concentration matters more than geographic proximity. These results inform policies on metropolitan logistics connectivity, related-industry collaboration, and long-term relational contracting.

Online social networks such as Twitter and Facebook are becoming popular form of social information networks. There are frequently many kinds of relationships in an online social network. Complex Network acting as one kind of big data technologies is often used to analyze users' social activities. By studying the Douban network, which is a representative multi-relationship online social network in China, big data of friendship relationship and book comments similar relationship are crawled through network topology measurement software, from the perspective of topological characteristics of complex network, the basic topologies of the two relationship networks constructed individually by the two relationships are analyzed. Based on these, a multi-relationship online social network based on Multi-subnet Composited Complex Network Model is constructed through loading book comments similar relationship subnet to follower relationship subnet, and accurate understanding of topologies of Douban multi-relationship network is obtained. These findings provide a deep understanding on the evolution of multi-relationship online social network, and can provide guidelines on how to build an efficient multi-relationship online social network evolution model.

The complex network community mining is a multidisciplinary research hotspot. It has been widely used in a terrorist organization to identify, protein function prediction, the metabolic pathways forecast, web community mining and link prediction in many areas. This article is mainly based on the genetic algorithm of the Community mining research oriented edge gene bit fast and effective local search variation algorithm. The algorithm uses a graph-based encoding strategy, the modules function as the objective function, Markov random walk method has certain the community divided accuracy and diversity of the initial population. The experiments show that the algorithm of the search space is effectively reduced, so the search efficiency can be further raised.

The information that caused by the complex network is massive, but because of a large amount of information, so the use of traditional data analysis has been unable to meet the search and mining complex network data, so the distributed computing mode, i.e., cloud computing has become the main method of calculation method of complex network, according to the calculation of cloud computing, it needs to consider computing topological partition method and the computational performance. This paper presents a data mining model matrix, according to this model can integrate different information, optimization of data mining, so as to improve the efficiency of complex network distributed computing.

A complex network is a interaction network of entities where global behavior is not deductible from the individual behaviors of each entities, leading to new properties emergence. Our problem is the network analysis ad modeling. Network analysis needs a formalism to assemble together the structure (static approach) and the function (dynamic approach), and to have a better understanding of the networks characteristics. In this paper, we introduce common used network modeling based on graph theory, having the role to simulate complex networks.

A new algorithm named Differential Evolution Algorithm for Community Detection (DEACD) was proposed in the paper. DEACD used DE as its search engine and used the network modularity as the fitness function to search for an optimal community partition of a network. In this algorithm, there is a modified binomial crossover mechanism to transmit some important information about the community structure in evolution effectively. In addition, a biased process and clean-up operation were employed in DEACD to improve the quality of the community partitions detected in evolution. Experimental results showed that DEACD has very competitive performance compared with other state-of-the-art community detection algorithms. In the process of evolution, the colony evolution was conducted under DE scheme, the network modularity was used to evaluate the fitness of individuals in the colony. The performance of DECD was analyzed by computer generated network and real-world network examples. The algorithm was implemented using matlab Genetic Algorithm Optimization Toolbox (GAOT), and the parametric analysis was performed in the experiment.

With appearance of small-world networks and scale-free networks, complex networks have provided natural theories to describe a variety of complex systems in many domains of science. In this paper, from the standpoint of Raman spectrum research aspect in material science, Raman spectrum analysis of graphite and diamond microcrystal retained by gallium phosphide nanoparticles is introduced. We found that molecular configuration of basic fuchsin adsorbed on gallium phosphide nanoparticles through experiments. Based on these, we adopt corresponding method of complex network to research some problems in Raman spectrum analysis and directly use complex network to convert real agglomerations into network model. With the intention of studying the topologies of agglomeration network, synchronizability behavior of agglomeration network are also discussed.