• Proceedings of the Korean Society of Computer Information Conference
• /
• 2014.01a
• /
• pp.9-13
• /
• 2014

The performance of large scale software applications has been automatically increasing for last few decades under the influence of Moore's law - the number of transistors on a microprocessor roughly doubled every eighteen months. However, on-chip transistors limitations and heating issues led to the emergence of multicore processors. The energy efficient ARM based System-on-Chip (SoC) processors are being considered for future high performance computing systems. In this paper, we present a case study of two widely used parallel programming models i.e. MPI and MapReduce on distributed memory cluster of ARM SoC development boards. The case study application, Black-Scholes option pricing equation, was parallelized and evaluated in terms of power consumption and throughput. The results show that the Hadoop implementation has low instantaneous power consumption that of MPI, but MPI outperforms Hadoop implementation by a factor of 1.46 in terms of total power consumption to execution time ratio.

• The KIPS Transactions:PartA
• /
• v.13A no.2 s.99
• /
• pp.157-162
• /
• 2006

Recent High Performance Computing (HPC) platforms can be classified as Shared-Memory Multiprocessors (SMP), Massively Parallel Processors (MPP), and Clusters of computing nodes. These platforms are deployed in many scientific and engineering applications which require very high demand on computing power. In order to realize an optimal performance for these applications, it is crucial to find and use the suitable computing platforms and programming paradigms. In this paper, we use SPEC HPC 2002 benchmark suite developed in various parallel programming models (MPI, OpenMP, and hybrid of MPI/OpenMP) to find an optimal computing environments and programming paradigms for them through their performance analyses.

• Bulletin of the Korean Chemical Society
• /
• v.21 no.10
• /
• pp.953-954
• /
• 2000

The equilibrium geometrical structures of silver atom clusters at their electronic ground states have been theo-retically determined by using the nonrelativistic semiempirical INDO/1 method. The clusters investigated are Agn, Agn+, and Agn- (n = 7 , 8, 9). In order to find the most stable structure, i.e., the global minimum in energy hypersurface, geometry optimization and energy calculation processes have been repeatedly performed for all the possible graphical models by changing the bond parameters (resonance integral values). The heptamers are pentagonal bipyramidal-Ag7(D5h), Ag7+ (D5h), Ag7- (D5h); the octamers are pentagonal bipyramidal with one atom capped-Ag8(D2d), Ag8+ (Cs), Ag8- (D2d); the nonamers are pentagonal bipyramidal with two atoms capped -Ag9(C2v), Ag9+ (C2v), Ag9- (C2v). Our structures are in good agreement with those by ab initio calculations ex-cept for the anionic Ag9- cluster. And it is noted that the INDO/1 method can accurately predict the Ag cluster geometries when a proper set of bond parameters is used.

• The Bulletin of The Korean Astronomical Society
• /
• v.46 no.2
• /
• pp.79.4-80
• /
• 2021

Sunspots' subsurface structure is an important subject to explain their stability and energy transport. Previous studies suggested two models for the subsurface structure of sunspots: monolithic model and cluster model. However, it is not revealed which model is more plausible so far. We obtain clues about the subsurface structure of sunspots by analyzing the motion of umbral flashes observed by the IRIS Mg II 2796Å slit-jaw images (SJI). The umbral flashes are believed as shock phenomena developed from upward propagating slow magnetohydrodynamic (MHD) waves. If the MHD waves are generated by convective motion below sunspots, the apparent origin of the umbral flashes known as oscillation center will indicate the horizontal position of convection cells. Thus, the distribution of the oscillation centers is useful to investigate the subsurface structure of sunspots. We analyze the spatial distribution of oscillation centers in the merged sunspot. As a result, we found that the oscillation centers distributed over the whole umbra regardless of the convergent interface between two merged sunspots. It implies that the subsurface structure of the sunspot is not much different from the convergent interface, and supports that many field-free gaps may exist below the umbra as the cluster model expected. For more concrete results, we should confirm that the oscillation centers determined by the umbral flashes accurately reflect the position of wave sources.

• Journal of Korean Society for Quality Management
• /
• v.51 no.4
• /
• pp.497-514
• /
• 2023

Purpose: This study aims to estimate the demand for various public parking lots in Seoul by clustering similar demand types of parking lots and predicting the demand for new public parking lots. Methods: We examined real-time parking information data and used time series clustering analysis to cluster public parking lots with similar demand patterns. We also performed various regression analyses of parking demand based on diverse heterogeneous data that affect parking demand and proposed a parking demand prediction model. Results: As a result of cluster analysis, 68 public parking lots in Seoul were clustered into four types with similar demand patterns. We also identified key variables impacting parking demand and obtained a precise model for predicting parking demands. Conclusion: The proposed prediction model can be used to improve the efficiency and publicity of public parking lots in Seoul, and can be used as a basis for constructing new public parking lots that meet the actual demand. Future research could include studies on demand estimation models for each type of parking lot, and studies on the impact of parking lot usage patterns on demand.

• The KIPS Transactions:PartB
• /
• v.14B no.2
• /
• pp.135-140
• /
• 2007

Feature selection is a preprocessing technique commonly used on high dimensional data. Feature selection studies how to select a subset or list of attributes that are used to construct models describing data. Feature selection methods attempt to explore data's intrinsic properties by employing statistics or information theory. The recent developments have involved approaches like correlation method, dimensionality reduction and mutual information technique. This feature selection have become the focus of much research in areas of applications with massive and complex data sets. In this paper, we provide a feature selection method considering data characteristics and generalization capability. It provides a computational approach for feature selection based on fuzzy cluster analysis of its attribute values and its performance measures. And we apply it to the system for classifying computer virus and compared with heuristic method using the contrast concept. Experimental result shows the proposed approach can give a feature ranking, select the features, and improve the system performance.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.6
• /
• pp.1239-1247
• /
• 2015

In this study, 64 administrative regions with high frequencies of apartment trade in Seoul, Korea are classified by the apartment sale price. To consider distributions of apartment price for each region as well as the mean of the price, the symbolic histogram-valued data approach is employed. Symbolic data include all types of data which have internal variation in themselves such as intervals, lists, histograms, distributions, and models, etc. As a result of the cluster analysis using symbolic histogram data, it is found that Gangnam, Seocho, and Songpa districts and regions near by those districts have relatively higher prices and larger dispersions. This result makes sense because those regions have good accessibility to downtown and educational environment.

• Journal of the Korean Chemical Society
• /
• v.40 no.4
• /
• pp.264-272
• /
• 1996

Cluster models of the Υ-Al2O3(111) and the Pt(111) surfaces have been used in an atom superposition and electron delocalization molecular orbital study of interfacial bond strengths between them. The reduced extents for Al3+ are due to the ratio of oxygen to aluminum atoms. The greater the reduced extent for Al3+ is, the stronger the binding energy is to Pt atoms in a cluster. The oxygen-covered surfaces of Υ-Al2O3(111) are shown to bind more weakly to Pt atoms, while the binding to the oxygen-covered surface formed under oxidizing conditions of Pt atoms is strong. The interfacial bond of platinum-alumina may be possible by a charge-transfer mechanism from the platinum surface to the partially empty O-2p band and Al3+ dangling surface orbital.

• Korean Journal of Materials Research
• /
• v.17 no.5
• /
• pp.250-255
• /
• 2007

The alloying effects on the electronic structures of Zinc are investigated using the relativistic $DV-X{\alpha}molecular$ orbital method in order to obtain useful information for alloy design. A new parameter which is the d obital energy level(Md) and the bonder order(Bo) of alloying elements in Zinc was introduced and used for prediction of the mechanical properties. The Md correlated with the atomic radius and the electronegativity of elements. The Bo is a measure of the strength of the covalent bond between M and X atoms. First-principles calculations of electronic structures were performed with a series of models composed of a MZn18 cluster and the electronic states were calculated by the discrete variational- $X{\alpha}method$ by using the program code SCAT. The central Zinc atom(M) in the cluster was replaced by various alloying elements. In this study energy level structures of pure Zinc and alloyed Zinc were calculated. From calculated results of energy level structures in MZn18 cluster, We found Md and Bo values for various elements of Zn. In this work, Md and Bo values correlated to the tensile strength for the Zn. These results will give some guide to design of zinc based alloys for high temperature applications and it is possible the excellent alloys design.

• The Korean Journal of Applied Statistics
• /
• v.30 no.1
• /
• pp.57-68
• /
• 2017

Clustering algorithms attempt to find a partition of a finite set of objects in to a potentially predetermined number of nonempty subsets. Gibbs sampling of a normal mixture of linear mixed regressions with a Dirichlet prior distribution calculates posterior probabilities when the number of clusters was known. Our approach provides simultaneous partitioning and parameter estimation with the computation of classification probabilities. A Monte Carlo study of curve estimation results showed that the model was useful for function estimation. Examples are given to show how these models perform on real data.