• Review of KIISC
• /
• v.33 no.2
• /
• pp.69-77
• /
• 2023

미국 국립표준기술연구소에서 개최한 양자내성암호 표준화 공모전이 2022년 표준화 선정을 마치고 4라운드 추가 표준선정에 진입하였다. 공모전에 제출된 후보 알고리즘은 각자의 성능 지표를 제시하였지만, 서로 다른 환경에서 성능을 측정하였기에 객관적인 비교가 어려웠다. 이를 해결하기 위해 양자내성암호 벤치마크 플랫폼이 등장하였다. 양자내성암호 벤치마크 플랫폼은 양자내성암호 후보 알고리즘을 모두 동일한 환경에서 벤치마크하여 객관적인 결과를 제공하는데 집중하였다. 이때 각 알고리즘이 지니는 다양한 구동 의존성 때문에 알고리즘을 동일 환경에서 가동하는 데에 제한이 있다. 따라서 벤치마크 플랫폼은 알고리즘들의 의존성을 제거하고 동일한 환경을 제공하는데 집중하였으며 이를 통해 양자내성암호의 성능에 대한 접근성을 높여 사람들의 양자내성암호 연구에 큰 도움을 주었다.

• 한국전산유체공학회:학술대회논문집
• /
• 1999.11a
• /
• pp.201-206
• /
• 1999

현재 전세계적으로 과한 기술 연산용 리눅스 클러스터에 대한 연구와 개발이 진행되고 있는 추세이다. 본 논문에서는 펜티엄 프로세서와 알파 프로세서 및 고속 이더넷을 이용한 리눅스 클러스터를 구현하고 이를 전산유체역학 분야에 적용하기 위한 선행연구로서 NAS의 병렬 벤치마크 프로그램인 NPB를 사용하여 리눅스 클러스터들의 성능을 비교 분석하였다. NPB 프로그램은 전산유체역학 분야의 해석프로그램에서 널리 이용되는 알고리즘들로 구성된 벤치마크의 프로그램으로 이를 이용한 성능 평가의 결과는 향후 리눅스 클러스터에서 실제 전산유체역학 코드들의 성능을 나타내는 중요한 지표가 될 수 있다. NPB 벤치마크 결과 리눅스 클러스터는 고가의 슈퍼컴퓨터에 비해 전산유체역학 분야에서 높은 가격 대 성능비를 보임을 알 수 있으며 이를 통하여 저비용 슈퍼컴퓨팅의 가능성을 제시하고자 한다.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.47 no.6
• /
• pp.57-67
• /
• 2010

This paper studies thermal characteristics of a mix of CPU-intensive and memory-intensive application workloads on a multi-core processor. Especially, we focus on thermal variations during program execution because thermal variations are more critical than average temperatures and their ranges for thermal management. New metrics are proposed to quantify such thermal variations for a workload. We study the thermal variations using SPEC CPU2006 benchmarks with varying cooling conditions and frequencies on an Intel Core 2 Duo processor. The results show that applications have distinct thermal variations characteristics. Such variations are affected by cooling conditions,operating frequencies and multiprogramming workload. Also, there are distinct spatial thermal variations between cores. Our new metrics and their results from this study provide useful insight for future research on multi-core thermal management.

• Journal of Korean Library and Information Science Society
• /
• v.35 no.2
• /
• pp.323-350
• /
• 2004

This study intends to offer the success factor in benchmarking technique and its application for library through investigation of benchmarking's theoretic background, such as concepts, implementation process, fitness models, and application case studies. It is expected that this study could guide to library fields and apply to basic data in next studies.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2014.07a
• /
• pp.1-4
• /
• 2014

인텔 Many Integrated Core (MIC) 아키텍쳐는 61개의 코어가 하나의 칩에 결합되어 있다. Xeon Phi 로 명명된 인텔 MIC는 인텔 E5 Xeon CPU 보다 2배의 single precision GFLOPs 성능을 제공한다. 인텔 MIC 는 수치연산에 최적화 되어 있는 아키텍쳐이다. 우리는 Xeon Phi 7120P를 가지고 벤치마킹을 하였고 클락스피드 1.238GHz, 61Core 이고 한 개의 코어당 4쓰레드를 사용하며 이론상 최고 성능은 Peak Double Precision(GFLOP)는 약 2-TFlops 이다. 이에 우리는 인텔 X86 아키텍쳐에서 openMP 와 인텔 MKL(Math kernel library) 라이브러리를 이용한 병렬프로그램을 작성하여 쓰레드 수를 증가 시키면서 인텔 Xeon Phi 와 E5 Xeon CPU에서 single precision 성능을 벤치마킹 하여, Xeon Phi 와 Xeon E5 의 이론적인 성능을 비교해 보고자 한다. 또한 openMP와 인텔 MKL라이브러리를 사용한 병렬환경에서 CPU의 성능 지표인 클락스피드와 코어수 외에 Vector unit size 의 크기가 성능에 어떤 영향을 미치는지 살펴보았다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2013.06a
• /
• pp.200-203
• /
• 2013

본 논문에서는 시장 점유율 상위를 차지하는 3 종의 해외 유명 제조사의 상용 HD 급 실시간 방송용 H.264 인코더의 화질 성능에 대해 전 기준 화질 평가를 통해 비교한다. 이를 위해 VQEG(Video Quality Experts Group)등에서 배포하는 9 종의 기준영상을 이용하여 PSNR(Peak Signal to Noise Ratio)을 측정하였으며 인코더 별 성능 비교를 위해 9 종의 기준영상에 대한 평균 PSNR 과 인코더/디코더의 전체 지연시간을 측정하였다. 그 결과 장비 별로 조금씩 다른 특성 결과와 현재 H.264 인코더 시험 시 성능 평가의 애매한 부분이었던 화질에 대한 벤치마크 결과를 얻을 수 있었다. 본 논문의 실험 결과는 국산 방송용 인코더의 객관적인 성능 평가를 위한 지표로 사용될 수 있을 것으로 예측된다.

• Land and Housing Review
• /
• v.11 no.2
• /
• pp.33-46
• /
• 2020

As the fastest growing office transaction volume in Korea, there's been a need for development of indicators to accurately diagnose the office capital market. The purpose of this paper is experimentally calculate to the office price index for effective benchmark indices in Seoul. The quantitative methodology used a Case-Shiller Repeat Sales Model (1991), based on actual multiple office transaction dataset with over minimum 1,653 ㎡ from Q3 1999 to 4Q 2019 in the case of 1,536 buildings within Seoul Metropolitan. In addition, the collected historical data and spatial statistical analysis tools were treated with the SAS 9.4 and ArcGIS 10.7 programs. The main empirical results of research are briefly summarized as follows; First, Seoul office price index was estimated to be 344.3 point (2001.1Q=100.0P) at the end of 2019, and has more than tripled over the past two decades. it means that the sales price of office per 3.3 ㎡ has consistently risen more than 12% every year since 2000, which is far above the indices for apartment housing index, announced by the MOLIT (2009). Second, between quarterly and annual office price index for the two-step estimation of the MIT Real Estate Research Center (MIT/CRE), T, L, AL variables have statistically significant coefficient (Beta) all of the mode l (p<0.01). Third, it was possible to produce a more stable office price index against the basic index by using the Moore-Penrose's pseoudo inverse technique at low transaction frequency. Fourth, as an lagging indicators, the office price index is closely related to key macroeconomic indicators, such as GDP(+), KOSPI(+), interest rates (5-year KTB, -). This facts indicate that long-term office investment tends to outperform other financial assets owing to high return and low risk pattern. In conclusion, these findings are practically meaningful to presenting an new office price index that increases accuracy and then attempting to preliminary applications for the case of Seoul. Moreover, it can provide sincerely useful benchmark about investing an office and predicting changes of the sales price among market participants (e.g. policy maker, investor, landlord, tenant, user) in the future.

• Korean Journal of Cognitive Science
• /
• v.34 no.3
• /
• pp.197-226
• /
• 2023

Generative Text Summarization is one of the Natural Language Processing tasks. It generates a short abbreviated summary while preserving the content of the long text. ROUGE is a widely used lexical-overlap based metric for text summarization models in generative summarization benchmarks. Although it shows very high performance, the studies report that 30% of the generated summary and the text are still inconsistent. This paper proposes a methodology for evaluating the performance of the summary model without using the correct summary. AggreFACT is a human-annotated dataset that classifies the types of errors in neural text summarization models. Among all the test candidates, the two cases, generation summary, and when errors occurred throughout the summary showed the highest correlation results. We observed that the proposed evaluation score showed a high correlation with models finetuned with BART and PEGASUS, which is pretrained with a large-scale Transformer structure.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.9
• /
• pp.1049-1059
• /
• 2019

We propose a learning-based image segmentation algorithm. Starting from super-pixels, our method learns the probability of merging two regions based on the ground truth made by humans. The learned information is used in determining whether the two regions should be merged or not in a segmentation stage. Unlike exiting learning-based algorithms, we use both local and object information. The local information represents features computed from super-pixels and the object information represent high level information available only in the learning process. The object information is considered as privileged information, and we can use a framework that utilize the privileged information such as SVM+. In experiments on the Berkeley Segmentation Dataset and Benchmark (BSDS 500) and PASCAL Visual Object Classes Challenge (VOC 2012) data set, out model exhibited the best performance with a relatively small training data set and also showed competitive results with a sufficiently large training data set.

• Journal of the Korean Geotechnical Society
• /
• v.38 no.11
• /
• pp.107-118
• /
• 2022

The social need for stable hydrogen storage technologies that respond to the increasing demand for hydrogen energy is increasing. Among them, underground hydrogen storage is recognized as the most economical and reasonable storage method because of its vast hydrogen storage capacity. In Korea, low-depth hydrogen storage using artificial protective structures is being considered. Further, establishing corresponding safety standards and ground stability evaluation is becoming essential. This study evaluated the hydro-mechanical behavior of the ground during a hydrogen gas leak from a low-depth underground hydrogen storage facility through the HM coupled analysis model. The predictive reliability of the simulation model was verified through benchmark experiments. A parameter study was performed using a metamodel to analyze the sensitivity of factors affecting the surface uplift caused by the upward infiltration of high-pressure hydrogen gas. Accordingly, it was confirmed that the elastic modulus of the ground was the largest. The simulation results are considered to be valuable primary data for evaluating the complex analysis of hydrogen gas explosions as well as hydrogen gas leaks in the future.