• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2020년도 제32회 한글 및 한국어 정보처리 학술대회
• /
• pp.93-97
• /
• 2020

본 연구는 딥 러닝 기반 의존 구문 분석에서, 학습에 적용하는 손실 함수에 따른 성능을 평가하였다. Pointer Network를 이용한 Left-To-Right 모델을 총 세 가지의 손실 함수(Maximize Golden Probability, Cross Entropy, Local Hinge)를 이용하여 학습시켰다. 그 결과 LH 손실 함수로 학습한 모델이 선행 연구와 같이 MGP 손실 함수로 학습한 것에 비해 UAS/LAS가 각각 0.86%p/0.87%p 상승하였으며, 특히 의존 거리가 먼 경우에 대하여 분석 성능이 크게 향상됨을 확인하였다. 딥러닝 의존 구문 분석기를 구현할 때 학습모델과 입력 표상뿐만 아니라 손실 함수 역시 중요하게 고려되어야 함을 보였다.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2019년도 제31회 한글 및 한국어 정보처리 학술대회
• /
• pp.543-545
• /
• 2019

본 논문은 최신 한국어 의존 구문 분석 모델(Korean dependency parsing model)들과 다양한 앙상블 모델(ensemble model)들을 결합하여 그 성능을 분석한다. 단어 표현은 미리 학습된 워드 임베딩 모델(word embedding model)과 ELMo(Embedding from Language Model), Bert(Bidirectional Encoder Representations from Transformer) 그리고 다양한 추가 자질들을 사용한다. 또한 사용된 의존 구문 분석 모델로는 Stack Pointer Network Model, Deep Biaffine Attention Parser와 Left to Right Pointer Parser를 이용한다. 최종적으로 각 모델의 분석 결과를 앙상블 모델인 Bagging 기법과 XGBoost(Extreme Gradient Boosting) 이용하여 최적의 모델을 제안한다.

• 한국전기전자재료학회:학술대회논문집
• /
• 한국전기전자재료학회 1996년도 춘계학술대회 논문집
• /
• pp.202-205
• /
• 1996

In this paper it is researched a relation between network structures and electrical properties - especially dielectric characteristics with changing of network structure. It is resulted that the specimens which have single network structures have smaller dielectric loss than SIN specimens but have relatively larger dependency to variation of temperature and frequency. For that reason formation of structures is attained by introducing of SIN to insulating materials. therefore it is counted that introduction of multiple structure including SIN is necessary to improve heat proof and electrical properties.

• 한국경영과학회:학술대회논문집
• /
• 한국경영과학회 2004년도 추계학술대회 및 정기총회
• /
• pp.644-647
• /
• 2004

Developing into information oriented society, dependency on network systems increases gradually in the area of economy, society and so on. Under that circumstance, temporary stop or damage of integrity of network system may have a strong impact on economic activities and overall society. Recently, threat of cyber terrorism, organized hostile action to government or industry through network, is increasing. Because of the features of network, it is not sufficient only a nation's countermeasure against Cyber terrorism, so international cooperation is needed. We review the status of countermeasure against cyber terrorism and examine several considerations for policy establishment.

• 환경정책연구
• /
• 제14권2호
• /
• pp.27-47
• /
• 2015

투입 산출변수의 외생적 흐름에 기초한 전통적인 자료포락분석은 생산활동의 모든 내부과정을 'black box'로 간주하며 각 생산단계의 상호작용에 대한 정보를 고려하지 못한 한계를 지닌다. 이에 생산과정을 세분화하고 각 생산단계의 효율성을 연계하여 추정하는 Network DEA 방식이 최근 활발히 적용되고 있으며 특히 환경효율성의 추정과정을 생산 및 저감활동으로 구분한 연구가 대표적이다. 본고는 한국의 51개 산업단지를 대상으로 Network DEA에 기초한 환경효율성 추정을 시도하였으며, 분석결과 추정치의 변별력, 비효율의 발생단계, 기타 외생변수와의 통계적 관련성 등의 관점에서 전통적 DEA 추정결과에 비해 우수한 점을 확인할 수 있었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권2호
• /
• pp.686-710
• /
• 2019

This work proposes a model to solve the problem of Network Coding over a one-session multicast network. The model is based on a system of restrictions that defines the packet flows received in the sink nodes as functions of the outgoing flows from the source node. A multicast network graph is used to derive a directed labeled line graph (DLLG). The successive powers of the DLLG adjacency matrix to the convergence in the null matrix permits the construction of the jump matrix Source-Sinks. In its reduced form, this shows the dependency of the incoming flows in the sink nodes as a function of the outgoing flows in the source node. The emerging packets for each outgoing link from the source node are marked with a tag that is a linear combination of variables that corresponds to powers of two. Restrictions are built based on the dependence of the outgoing and incoming flows and the packet tags as variables. The linear independence of the incoming flows to the sink nodes is mandatory. The method is novel because the solution is independent of the Galois field size where the packet contents are defined.

• 한국전산유체공학회:학술대회논문집
• /
• 한국전산유체공학회 2001년도 춘계 학술대회논문집
• /
• pp.149-154
• /
• 2001

Parallel performance of a Myrinet based PC-cluster was tested and compared with a conventional Fast-Ethernet system. A preconditioned Navier-Stokes code was parallelized with domain decomposition technique, and used for the parallel performance test. Speed-up ratio was examined as a major performance parameter depending on the number of processor and the network topology. As was expected, Myrinet system shows a superior parallel performance to the Fast-Ethernet system even with a single network adpater for a dual processor SMP machine. A test for the dependency on problem size also shows that network communication speed is a crucial factor for parallelized computational fluid dynamics analysis and the Myrinet system is a plausible candidate for high performance parallel computing system.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 1995년도 추계학술대회 논문집
• /
• pp.697-701
• /
• 1995

The network encironment includes a number of autonomous agents which are widely distributed, have different platforms, and change very dynamically. The information system operated on this environment must solve some basic problems; restrictive client-server models, heterogeneous systems, and intellignet agents. We are using KQML for knowledge management. Java language provides solutions against a strong server dependency and a heterogenous troubles. We introduce KQML as an agent communication language and JATLite as a java agent template. these increase an efficiency of communication on network and enable us to use resources distributed in world wide web. Also, we describe a new agent system architecture and implement it through an example scenario

• 한국IT서비스학회지
• /
• 제7권1호
• /
• pp.195-204
• /
• 2008

As the dependency to network system and demands of efficient storage systems rapidly grows in every networking filed, the current trends initiated by explosive networked data grow due to the wide-spread of internet multimedia data and internet requires a paradigm shift from computing-centric to data-centric in storage systems. Furthermore, the new environment of file systems such as SAN(Storage Area Network) is adopted to the existing storage paradigm for providing high availability and efficient data access. We describe the design issues and system components of $SANique^{TM}$, which is the cluster file system based on SAN environment. We, especially, present the comparative results of performance analysis for the intensive I/O test by using the DBMSs that are operated at the top of cluster file system $SANique^{TM}$, EXT3 and NFS respectively.

• 한국항해항만학회지
• /
• 제47권2호
• /
• pp.93-99
• /
• 2023

This study aimed to determine relationships among risk factors influencing container port operation using Bayesian network. Risk factors identified from prior studies were classified into five groups: human error, machinery error, environmental risk, security risk, and natural disasters. P anel experts discussed identified risk factors to fulfil conditional probability tables of the interdependence model. The interdependence model was also validated by sensitivity analysis and provided an interrelation of factors influencing the direction of each other. Results of the interdependence model were partially in line with results from prior studies while practices in the global port industry confirmed interrelationships of risk factors. In addition, the relationship between top-ranked risk factors can provide a schematic drawing of the model. Accordingly, results of this study can expand the prior research in the Korean port industry, which may help port authorities improve risk management and reduce losses from the risk.