• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.127-146
• /
• 2022

Recently, as word embedding has shown excellent performance in various tasks of deep learning-based natural language processing, researches on the advancement and application of word, sentence, and document embedding are being actively conducted. Among them, cross-language transfer, which enables semantic exchange between different languages, is growing simultaneously with the development of embedding models. Academia's interests in vector alignment are growing with the expectation that it can be applied to various embedding-based analysis. In particular, vector alignment is expected to be applied to mapping between specialized domains and generalized domains. In other words, it is expected that it will be possible to map the vocabulary of specialized fields such as R&D, medicine, and law into the space of the pre-trained language model learned with huge volume of general-purpose documents, or provide a clue for mapping vocabulary between mutually different specialized fields. However, since linear-based vector alignment which has been mainly studied in academia basically assumes statistical linearity, it tends to simplify the vector space. This essentially assumes that different types of vector spaces are geometrically similar, which yields a limitation that it causes inevitable distortion in the alignment process. To overcome this limitation, we propose a deep learning-based vector alignment methodology that effectively learns the nonlinearity of data. The proposed methodology consists of sequential learning of a skip-connected autoencoder and a regression model to align the specialized word embedding expressed in each space to the general embedding space. Finally, through the inference of the two trained models, the specialized vocabulary can be aligned in the general space. To verify the performance of the proposed methodology, an experiment was performed on a total of 77,578 documents in the field of 'health care' among national R&D tasks performed from 2011 to 2020. As a result, it was confirmed that the proposed methodology showed superior performance in terms of cosine similarity compared to the existing linear vector alignment.

• Journal of The Korean Association of Information Education
• /
• v.13 no.2
• /
• pp.225-236
• /
• 2009

According to the revised guideline for information & communication technology, algorithm has to be dealt with the learning contents from elementary school, so now it needs to study various educational methods for students to learn algorithm easily and effectively. In this study, after selecting sort algorithms as the content of learning, which is recognized as an important part of algorithm, an playing activity-based algorithm teaching method and an animation-based algorithm learning and teaching method were applied for the lower graders of an elementary school. Sorting algorithm education is adopted in two different ways, then we got two conclusions after analyzing the results of a specially designed achievement test and the questionnaire. First, there were not great differences in both educational methods in academic achievement, but it was clear that the lower grade elementary students can learn algorithm with ease. Second, the playing activity-centered algorithm education is more effective than animation-based education in improving students' comprehension, interest and satisfaction.

• Journal of KIISE:Computer Systems and Theory
• /
• v.29 no.10
• /
• pp.550-560
• /
• 2002

External sort on cluster computers requires not only fast internal sorting computation but also careful scheduling of disk input and output and interprocessor communication through networks. This is because the overall time for the execution is determined by reflecting the times for all the jobs involved, and the portion for interprocessor communication and disk I/O operations is significant. In this paper, we improve the sorting performance (sorting throughput) on a cluster of PCs with a low-speed network by developing a new algorithm that enables even distribution of load among processors, and optimizes the disk read and write operations with other computation/communication activities during the sort. Experimental results support the effectiveness of the algorithm. We observe the algorithm reduces the sort time by 45% compared to the previous NOW-sort[1], and provides more scalability in the expansion of the computing nodes of the cluster as well.

• Journal of computational fluids engineering
• /
• v.19 no.1
• /
• pp.27-33
• /
• 2014

A new and efficient implicit scheme is proposed to obtain a steady-state solution in time integration and the comparison of characteristics with the approximation ways for the implicit method to solve the incompressible Navier-Stokes equations is provided. The conservative, finite-volume cell-vertex upwind scheme and artificial compressibility method using dual time stepping for time accuracy is applied in this paper. The numerical results obtained indicate that the direct application of Jacobian matrix to the Lower and upper sweeps of implicit LU-SGS leads to better performance as well as convergence regardless of CFL number and true time step than explicit scheme and approximation of Jacobian matrix. The flow simulation around box in uniform flow with unstructured meshes is demonstrated to check the validity of the current formulation.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37B no.10
• /
• pp.956-964
• /
• 2012

An opportunistic interference alignment based on dynamic cell selection is proposed. Since the proposed method can switch the desired signal space and the interference space of received signals, an additional selective diversity gain increases. The performance analysis using probabilistic models provides a mathematical expression for the sum-rate capacity. Simulation examples show that the proposed method achieves the higher sum-rate capacity than that of the conventional opportunistic interference alignment.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2010.08a
• /
• pp.269-269
• /
• 2010

$SnO_2$ 나노선은 n-type의 전기적 특성과 우수한 광 특성을 보이며, 전자소자, 광소자 뿐 아니라 다양한 종류의 가스 센서 등에 응용되고 있다. 그러나 $SnO_2$ 나노선은 공기중에서 전기적으로 불안정한 특성을 보이며, 도핑을 하지 않은 나노선 소자에서는 전자의 모빌러티가 높지 않다는 단점을 갖고 있다. 이를 개선하고자 본 연구에서는 화학기상증착법 (Chemical Vapor Deposition)으로 Sb을 도핑한 $SnO_2$ 나노선을 성장하여 전계방출효과 트랜지스터 (field effect transistor: FET)를 제작하여 전기적 특성과 UV 반응성의 변화를 측정하였다. Sb 도핑 양을 늘려감에 따라 전기적 특성이 반도체 특성에서 점점 금속 특성으로 변하는 것과 게이트 전압의 영향을 적게 받는 것을 확인하였다. 또한 도핑을 해준 $SnO_2$ 나노선의 경우 UV 반응과 회복 시간이 기존에 비하여 크게 감소하여 UV 센서에 더욱 적합해진 것을 확인하였다. 또한, 슬라이딩 트랜스퍼 공정을 이용하여 나노선을 원하는 기판에 정렬된 상태로 전이할 때 도핑한 나노선은 표면특성의 변화로 정렬도가 크게 감소하는 것을 확인하였고, 기판에 윤활제를 사용하여 정렬도를 높일 수 있었다.

• ETRI Journal
• /
• v.14 no.2
• /
• pp.120-127
• /
• 1992

슬리브내에서 페룰을 통해 광섬유의 정렬이 이루어지는 광커넥터의 대부분의 접속구조에서 각 구성품간의 어긋남 비정렬요인이 중복되어 나타나게 되므로 접속손실이 매우 큰 영향을 받고 있다. 이러한 접속손실의 향상을 위해 선진 외국에서는 각 구성품의 정밀도 향상과 함께 구성품의 저손실화 조립연구도 수행하고 있다. 본 연구에서는 광섬유 출사면에서의 최대 광도부분을 회전조정하여 최소영역에서 두 광섬유가 정렬, 접속되게 하는 저손실화 조립 장치 개발과 관련하여, 임의로 조립된 상태와 본 연구에서 제안한 적응 제어 알고리즘 (Adaptive Control Algorithm)을 이용하여 광Intensity의 최대부분을 회전조정한 상태와의 삽입손실의 통계적 분포를 비교.분석하였다.

• Annual Conference of KIPS
• /
• 2022.05a
• /
• pp.320-321
• /
• 2022

부두 내 컨테이너를 적재하는 과정에서 정렬 상태가 부정확한 경우 강풍으로 인한 안전사고가 발생할 가능성이 있다. 본 논문에서는 컨테이너 안전사고를 예방하기 위한 딥러닝 기반의 컨테이너 정렬 상태 분류 알고리즘을 제안한다. 제안하는 알고리즘은 정렬을 분류하는 기준을 제시하고 YOLO 기반의 모델을 구현했다. 추론 속도, 검출 정확도, 분류 정확도를 기준으로 각 모델의 성능을 평가했으며 성능 결과는 YOLOv4모델이 YOLOv3모델에 비해서 추론 속도는 느리지만, 검출 정확도와 분류 정확도는 높음을 보인다.

• Survey Research
• /
• v.6 no.2
• /
• pp.1-32
• /
• 2005

The accuracy of exit poll mainly depends on the sampling method of voting places. For exit poll, we propose a probability sampling method of selecting voting places as an alternative to the bellwether polling place sampling. Through an empirical study based on the 2004 general election data, the efficiency of the suggested systematic sampling from ordered voting places was evaluated in terms of mean prediction error and it turns out that the proposed sampling method outperformed the bellwether polling places sampling. We also calculated the variance of estimator from the proposed sampling, and considered the sample size problem to guarantee the target precision using the design effect of the proposed sample design.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.10 no.8
• /
• pp.1843-1849
• /
• 2009

In real-time 3D Graphics, pipeline optimization is one of techniques enhancing rendering performance. Pipeline optimization is kind of buffer reordering problem, but it is NP-hard. Therefore techniques that is approximating optimal solution and suitable for real-time 3D graphics are needed. This paper analyze pattern of rendering states changing costs for real-time 3D graphics, and based on this, the algorithm that brings rendering states into line by changing costs is proposed. The proposed technique shows good performance enhancement when costs of some rendering states are much higher than others. Proposed technique shows 2.5 to 4 times better performance than non-ordering algorithm and becomes more faster when rendering costs of a state gets higher.