• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04a
• /
• pp.10-12
• /
• 2003

SMP 클러스터가 고성능 계산을 위한 플랫폼으로 등장함에 따라, 이 시스템을 활용하기 위한 프로그래밍 환경에 대한 관심이 증가하고 있다. 이 논문에서 우리는 ParADE라고 부르는 쉽고, 이식성이 높으며. 고성능의 프로그래밍이 가능한 새로운 프로그래밍 환경을 소개한다. ParADE는 OpenMP 프로그래밍 환경으로 HLRC 변종 프로토콜을 구현한 다중 쓰레드 DSM 시스템을 기반으로 하고 있다. 특별히. 이 논문에서는 성능 개선을 위한 OpenMP 변환기의 역할에 중점을 둔다. OpenMP 변화기는 OpenMP 프로그램 모델과 실행 시스템의 수행 모델 사이에서 가교 역할을 한다. 특히, OpenMP 변환기는 동기화 디렉티브를 변환하고 임계 영역에 있는 작은 변수의 메모리 일관성을 유지하기 위해 집합 통신 함수를 활용한다. 동기화 디렉티브 성능 측정을 위한 마이크로벤치마크 프로그램을 통한 실험에서 ParADE 시스템은 기존의 DSM 시스템에 비해 우수한 성능을 보였다.

• Electronics and Telecommunications Trends
• /
• v.33 no.4
• /
• pp.1-13
• /
• 2018

Deep learning (DL) models have been widely applied to AI applications such image recognition and language translation with big data. Recently, DL models have becomes larger and more complicated, and have merged together. For the accelerated training of a large-scale deep learning model, model parallelism that partitions the model parameters for non-shared parallel access and updates across multiple machines was provided by a few distributed deep learning frameworks. Model parallelism as a training acceleration method, however, is not as commonly used as data parallelism owing to the difficulty of efficient model parallelism. This paper provides a comprehensive survey of the state of the art in model parallelism by comparing the implementation technologies in several deep learning frameworks that support model parallelism, and suggests a future research directions for improving model parallelism technology.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.2
• /
• pp.37-46
• /
• 1996

This paper proposes a cost-effective solution to the synonym problem of virtual caches. In the proposed solution, a minimal hardware addition guarantees the correctness whereas the software counterpart helps improve the performance. The key to this proposed solution is an addition of a small physically-indexed cache called U-cache. The U-cache maintains the reverse translation information of the cache blocks that belong to unaligned virtual pages only, where aligned measns that the lower bits of the virtual page number match those of the corresponding physical page number. The page alignment is a simple software optimization to improve the performance of the U-cche hardware. With the combination of both hardware and software, the proposed solution reduces the hardware costs and minimizes software modification and performance degradation. Performance evaluation base on ATUM traces shows that a U-cache, with only a few entries, performs almost as well as fully-configured hardware-based solution when more than 95% of the pages are aligned.

• Journal of Communications and Networks
• /
• v.7 no.4
• /
• pp.499-508
• /
• 2005

In this paper, we address the problem of efficient routing and wavelength assignment (RWA) in multi-fiber wavelength division multiplexing (WDM) networks without wavelength translation, under dynamic traffic. We couple Dijkstra's shortest path algorithm with a suitable weight function which chooses optical paths based both on wavelength availability and multi-fiber segments. We compare our approach with other RWA schemes both for regular and irregular WDM multi-fiber network topologies in terms of blocking probability and overall link utilization.

• Journal of Electrical Engineering and information Science
• /
• v.3 no.1
• /
• pp.1-7
• /
• 1998

High-Level synthesis of digital circuits calls for automatic translation of a behavioral description to a structural design entity represented in terms of components and connection. One of the critical steps in high-level synthesis is to determine a particular scheduling algorithm that will assign behavioral operations to control states. A new scheduling algorithm called Data Availability Scheduling (DAS) for high-level synthesis is presented. It can determine an appropriate scheduling algorithm and minimize the number of states required using data availability and dependency conditions extracted from the behavioral code, taking into account of states required using data availability and dependency conditions extracted from the behavioral code, taking into account resource constraint in each control state. The DAS algorithm is efficient because data availability conditions, and conditional and wait statements break the behavioral code into manageable pieces which are analyzed independently. The output is the number of states in a finite state machine and shows better results than those of previous algorithms.

• ETRI Journal
• /
• v.25 no.4
• /
• pp.219-230
• /
• 2003

This paper describes a rule-based approach for syntactic disambiguation used by the English sentence parser in E-TRAN 2001, an English-Korean machine translation system. We propose Parser's Ambiguity Type Information (PATI) to automatically identify the types of ambiguities observed in competing candidate trees produced by the parser and synthesize the types into a formal representation. PATI provides an efficient way of encoding knowledge into grammar rules and calculating rule preference scores from a relatively small training corpus. In the overall scoring scheme for sorting the candidate trees, the rule preference scores are combined with other preference functions that are based on statistical information. We compare the enhanced grammar with the initial one in terms of the amount of ambiguity. The experimental results show that the rule preference scores could significantly increase the accuracy of ambiguity resolution.

• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.187-192
• /
• 2007

The purpose of this paper is to show that the current notation of Japanese proper names in Korean has some problems. It cannot represent the different sounds between the voiced and voiceless. The purpose of this paper is also to give a more correct notation which is coherent and efficient. After introducing some general knowledge about the phonemes of Japanese language, I measured the Voice Onset Time of the stops[k, t] at the beginning, in the middle and at the end of a word, and compared the spectrogram of affricates with that of fricatives. In conclusion, Japanese voiceless [k, t ,$t{\int}$] should be written as [ㅋ,ㅌ,ㅊ] and voiced [g, d $d_3$] as [ㄱ,ㄷ,ㅈ] and the affricate[ts] as[ㅊ] in Korean.

• Annual Conference on Human and Language Technology
• /
• 2005.10a
• /
• pp.89-96
• /
• 2005

본 연구에서는 영한 기계 번역에서 20단어 이상의 긴 문장을 보다 정확히 분석하기 위하여 문장을 복수개의 의미 있는 절로 분할하고자 한다. 긴 문장은 구문 분석을 시도할 때, 시간적으로 또는 공간적으로 급격히 증가하는 자원을 소모시킨다. 이러한 문제를 해결하기 위하여, 본 연구에서는 긴 문장에서 분할 가능한 지점을 인식하여 이러한 지점을 중심으로 여러 개의 절을 생성한 후, 이 절을 개별적으로 분석하고자 하였다. 문장을 분할하기 위해서 일단 문장 내부에 존재하고 있는 분할이 가능한 지점을 선택하고, 선택된 지점을 중심으로 문맥 정보를 표현하는 입력 벡터를 생성하였다. 그리고 Support Vector Machine (SVM)을 이용하여 이러한 후보 지점의 특성을 학습하여 향후 긴 문장이 입력되었을 때 보다 정확하게 분할점을 찾고자 하였다. 본 논문에서는 SVM의 보다 좋은 학습과 분류를 위하여 내부 커널로써 다항 커널 (polynomial kernel)을 사용하였다. 그리고 실험을 통하여 약 0.97의 f-measure 값을 얻을 수 있었다.

• Annual Conference on Human and Language Technology
• /
• 2001.10d
• /
• pp.288-293
• /
• 2001

한영 기계번역 시스템을 구현하기 위해서는 다양하게 활용하는 한국어 용언을 보다 효율적으로 처리해야 할 필요가 있다. 한국어 용언은 그 활용이 매우 다양하여 활용에 따라 문장 내에서 다양하게 기능하게 된다. 한영 기계번역 시스템에서는 용언의 활용이 가지는 여러 정보를 효율적으로 분석하여 해당정보를 보다 효과적으로 역문에 반영시키는 연구가 필요하다. 본 논문에서는 용언의 활용에 따른 여러 정보-시제에 관한 정보(선어말어미 관련), 문종에 관한 정보(어말어미 관련), 양상에 관한 정보(보조용언, 어말어미 관련) 등-를 통일된 코드를 이용하여 일괄적으로 처리하는 방법론과 그 과정을 제시한다.

• Journal of Korea Multimedia Society
• /
• v.17 no.4
• /
• pp.466-477
• /
• 2014

Many image hiding schemes based on least significant bit (LSB) transformation have been proposed. One of the LSB-based image hiding schemes that employs diamond encoding was proposed in 2008. In this scheme, the binary secret data is converted into base n representation, and the converted secret data is concealed in the cover image. Here, we show that this scheme has two vulnerabilities: noticeable spots in the stego-image, i.e., a non-smooth embedding result, and inefficiency caused by rough re-adjustment of falling-off-boundary value and impractical base translation. Moreover, we propose a new scheme that is efficient and produces a smooth and high quality embedding result by restricting n to power of 2 and using a sophisticated re-adjustment procedure. Our experimental results show that our scheme yields high quality stego-images and is secure against RS detection attack.