• 한국멀티미디어학회논문지
• /
• 제21권8호
• /
• pp.952-959
• /
• 2018

Clustering is one of the most fundamental algorithms in machine learning. The performance of clustering is affected by the distribution of data, and when there are more data or more dimensions, the performance is degraded. For this reason, we use a stacked auto encoder, one of the deep learning algorithms, to reduce the dimension of data which generate a feature vector that best represents the input data. We use k-means, which is a famous algorithm, as a clustering. Sine the feature vector which reduced dimensions are also multi dimensional, we use the Euclidean distance as well as the cosine similarity to increase the performance which calculating the similarity between the center of the cluster and the data as a vector. A deep clustering networks combining a stacked auto encoder and k-means re-trains the networks when the k-means result changes. When re-training the networks, the loss function of the stacked auto encoder and the loss function of the k-means are combined to improve the performance and the stability of the network. Experiments of benchmark image ad document dataset empirically validated the power of the proposed algorithm.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2008년도 추계학술발표대회
• /
• pp.332-335
• /
• 2008

In this paper, based on H-tree, which is proposed as the basic data cube structure for multi-dimensional data stream analysis, we have done some analysis. We find there are a lot of redundant nodes in H-tree, and the tree-build method can be improved for saving not only memory, but also time used for inserting tuples. Also, to facilitate more fast and large amount of data stream analysis, which is very important for stream research, H^*-tree is designed and developed. Our performance study compare the proposed H^*-tree and H-tree, identify that H^*-tree can save more memory and time during inserting data stream tuples.

• 대한기계학회논문집B
• /
• 제22권8호
• /
• pp.1173-1182
• /
• 1998

A throughflow analysis program for axial flow turbines is constructed, which can handle not only the two-dimensional multi-streamline (streamline curvature) method but also the one-dimensional mean line method. Calculations are performed for single stage and multi-stage axial flowturbines. For a wide operating range, the performance and flow field calculated by the present streamline curvature method are close enough to the test data. It is also revealed for the single stage turbine that the present analysis leads to far better correspondence with the experiment than other researchers" throughflow analyses. A special focus is put on the comparison of the results between the streamline curvature analysis and the mean line analysis. It is found that the mean line analysis can not predict the performance for highly off-designed conditions as accurately as the streamline curvature method, which shows the importance of considering the spanwise variation of loss and flow.

• Journal of Mechanical Science and Technology
• /
• 제19권8호
• /
• pp.1568-1575
• /
• 2005

The calculation of the transmission loss of the silencers with complicated internal structures by the conventional BEM combined with the transfer matrix method is incorrect at best or impossible for 3-dimensional silencers due to its inherent plane wave assumption. On this consideration, we propose an efficient practical means to formulate algebraic overall condensed acoustic equations for the whole acoustic structure, where particle velocities on the domain interface boundaries are unknowns, and the solutions are used later to compute the overall transfer matrix elements, based on the multi-domain BEM data. The transmission loss estimation by the proposed method is tested by comparison with the experimental one on an air suction silencer with perforated internal structures installed in air compressors. The method shows its viability by presenting the reasonably consistent anticipation of the experimental result.

• 한국컴퓨터정보학회논문지
• /
• 제19권9호
• /
• pp.21-31
• /
• 2014

특이치 분해는 다양한 분야의 데이터 집단에서 고유한 특성을 찾는 특징 추출 분야에 많이 활용되고 있다. 하지만 특이치 분해의 복잡 행렬 연산은 많은 연산 시간을 요구한다. 본 논문에서는 특이치 분해의 대표적인 알고리즘인 one-sided block Jacobi를 고속 처리하기 위해 2차원 멀티코어 시스템을 이용하여 효율적으로 병렬 구현하고 성능을 향상시킨다. 또한, one-sided block Jacobi 알고리즘의 다양한 행렬 ($128{\times}128$, $64{\times}64$, $32{\times}32$, $16{\times}16$)을 서로 다른 2차원 PE 구조에 구현하고 성능 및 에너지를 분석함으로써 각 행렬에 대한 최적의 멀티코어 구조를 탐색한다. 더불어 동일한 행렬의 one-sided block Jacobi 알고리즘에 대해 선택된 멀티코어 구조와 상용 고성능 그래픽스 프로세싱 유닛 (GPU)과의 성능 비교를 통해 제안한 2차원 멀티코어 방법의 잠재 가능성을 확인한다.

• 한국원자력학회:학술대회논문집
• /
• 한국원자력학회 2004년도 추계학술발표회 발표논문집
• /
• pp.363-364
• /
• 2004

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제5권12호
• /
• pp.2374-2391
• /
• 2011

A variety of reconstruction methods has been developed to convert a set of scattered points generated from real models into explicit forms, such as polygonal meshes, parametric or implicit surfaces. In this paper, we present a method to construct multi-scale implicit surfaces from scattered points using multiscale kernels based on kernel and multi-resolution analysis theories. Our approach differs from other methods in that multi-scale reconstruction can be done without additional manipulation on input data, calculated functions support level of detail representation, and it can be naturally expanded for n-dimensional data. The method also works well with point-sets that are noisy or not uniformly distributed. We show features and performances of the proposed method via experimental results for various data sets.

• International Journal of Quality Innovation
• /
• 제3권2호
• /
• pp.113-131
• /
• 2002

The relationships among multi-dimensional data (such as medical examination data) with ambiguity and variation are difficult to explore. The traditional approach to building a data classification system requires the formulation of rules by which the input data can be analyzed. The formulation of such rules is very difficult with large sets of input data. This paper first describes two classification approaches using back-propagation (BP) neural network and Mahalanobis distance (MD) classifier, and then proposes two classification approaches for multi-dimensional feature selection. The first one proposed is a feature selection procedure from the trained back-propagation (BP) neural network. The basic idea of this procedure is to compare the multiplication weights between input and hidden layer and hidden and output layer. In order to simplify the structure, only the multiplication weights of large absolute values are used. The second approach is Mahalanobis-Taguchi system (MTS) originally suggested by Dr. Taguchi. The MTS performs Taguchi's fractional factorial design based on the Mahalanobis distance as a performance metric. We combine the automatic thresholding with MD: it can deal with a reduced model, which is the focus of this paper In this work, two case studies will be used as examples to compare and discuss the complete and reduced models employing BP neural network and MD classifier. The implementation results show that proposed approaches are effective and powerful for the classification.

• 한국정보처리학회논문지
• /
• 제6권12호
• /
• pp.3469-3476
• /
• 1999

공간적인 자료구조를 기반으로 하는 공간 데이터베이스에서는 일차원 색인구조와는 달리 공간객체들의 다차원적인 특성에 부합되는 새로운 색인구조가 요구되고 있다. 본 논문에서는 이러한 요구사항을 충족시키기 위하여 기존 다차원 색인구조들의 특징 분석을 통하여 공간 데이터베이스의 효율적인 검색을 위한 새로운 색인구조를 제안하였다. 기존 X-트리에서 슈퍼노드의 순차적인 검색방법의 개선과 방대한 슈퍼노드가 생성되는 경우에도 검색시간의 단축이 가능하도록 하기 위하여, 포인트 색인구조를 갖는 kd-트리를 X-트리에 병합시킨 색인구조를 제안하였다. 제안된 색인구조를 실제로 구현하여 실험 데이터의 차원과 분포에 따라 검색시간을 분석하였다.

• 한국대기환경학회지
• /
• 제24권4호
• /
• pp.423-438
• /
• 2008

BEIS2 (Biogenic Emissions Inventory System version 2) and BEIS3.12 (BEIS version 3.12) were used to estimate hourly biogenic emissions over South Korea using a set of vegetation and meteorological data simulated with the MM5 (Mesoscale Model version 5). Two biogenic emission models utilized different emission factors and showed different responses to solar radiations, resulting in about $10{\sim}20%$ difference in the nationwide isoprene emission estimates. Among the 11-vegetation classes, it was found that mixed forest and deciduous forest are the most important vegetation classes producing isoprene emissions over South Korea comprising ${\sim}90%$ of the total. The simulated isoprene concentrations over Seoul metropolitan area show that diurnal and daily variations match relatively well with the PAMS (Photochemical Air Monitoring Station) measurements during the period of June 3${\sim}$June 10, 2004. Compared to BEIS2, BEIS3.12 yielded ${\sim}35%$ higher isoprene concentrations during daytime and presented better matches to the high peaks observed over the Seoul area. This study showed that the importance of vegetation data and emission factors to estimate biogenic emissions. Thus, it is expected to improve domestic vegetation categories and emission factors in order to better represent biogenic emissions over South Korea.