• The Journal of the Acoustical Society of Korea
• /
• v.39 no.6
• /
• pp.505-514
• /
• 2020

Since a large amount of training data is typically needed to train Deep Neural Networks (DNNs), a parallel training approach is required to train the DNNs. The Stochastic Gradient Descent (SGD) algorithm is one of the most widely used methods to train the DNNs. However, since the SGD is an inherently sequential process, it requires some sort of approximation schemes to parallelize the SGD algorithm. In this paper, we review various efforts on parallelizing the SGD algorithm, and analyze the computational overhead, communication overhead, and the effects of the approximations.

• Journal of Information Processing Systems
• /
• v.12 no.1
• /
• pp.35-45
• /
• 2016

To resolve the problems of Poisson/impulse noise, blurriness, and sharpness in degraded X-ray images, a novel and efficient enhancement algorithm based on X-ray image fusion using a discrete wavelet transform is proposed in this paper. The proposed algorithm consists of two basics. First, it applies the techniques of boundary division to detect Poisson and impulse noise corrupted pixels and then uses the Wiener filter approach to restore those corrupted pixels. Second, it applies the sharpening technique to the same degraded X-ray image. Thus, it has two source X-ray images, which individually preserve the enhancement effects. The details and approximations of these sources X-ray images are fused via different fusion rules in the wavelet domain. The results of the experiment show that the proposed algorithm successfully combines the merits of the Wiener filter and sharpening and achieves a significant proficiency in the enhancement of degraded X-ray images exhibiting Poisson noise, blurriness, and edge details.

• Journal of Korea Society of Industrial Information Systems
• /
• v.21 no.6
• /
• pp.13-21
• /
• 2016

Feature selection is the one of important issue in the field of data mining and machine learning. It is the technique to find a subset of features which provides the best classification performance, from the source data. We propose a feature subset selection method using the neighborhood rough set model based on information granularity. To demonstrate the effectiveness of proposed method, it was applied to select the useful features associated with breast tumor diagnosis of 298 shape features extracted from 5,252 breast ultrasound images, which include 2,745 benign and 2,507 malignant cases. Experimental results showed that 19 diagnostic features were strong predictors of breast cancer diagnosis and then average classification accuracy was 97.6%.

• Structural Engineering and Mechanics
• /
• v.67 no.5
• /
• pp.427-437
• /
• 2018

Structural pounding is commonly observed phenomenon during major ground motion, which can cause both structural and architectural damages. To reduce the amount of damage from pounding, the best and effective way is to increase the separation distance. Generally, existing design procedures for determining the separation distance between adjacent buildings subjected to structural pounding are based on approximations of the buildings' peak relative displacement. These procedures are based on unknown safety levels. The aim of this research is to estimate probabilistic separation distance between adjacent structures by considering the variability in the system and uncertainties in the earthquakes characteristics through comprehensive numerical simulations. A large number of models were generated using a robust Monte-Carlo simulation. In total, 6.54 million time-history analyses were performed over the adopted models using an ensemble of 25 ground motions as seismic input within OpenSees software. The results show that a gap size of 50%, 70% and 100% of the considered design code for the structural periods in the range of 0.1-0.5 s, leads to have the probability of pounding about 41.5%, 18% and 5.8%, respectively. Finally, based on the results, two equations are developed for probabilistic determination of needed structural separation distance.

• Journal of the Society of Naval Architects of Korea
• /
• v.42 no.4 s.142
• /
• pp.330-340
• /
• 2005

Compressible two-phase flow is analyzed based on the arbitrary Lagrangian-Eulerian (ALE) formulation. For water, Tamman type stiffened equation of state is used. Numerical fluxes are calculated using the ALE two-phase Godunov scheme which assumes only that the speed of sound and pressure can be provided whenever density and internal energy are given. Effects of the approximations of a material interface speed are Investigated h method Is suggested to assign a rigid body boundary condition effectively To validate the developed code, several well-known problems are calculated and the results are compared with analytic or other numerical solutions including a single material Sod shock tube problem and a gas/water shock tube problem The code is applied to analyze the refraction and transmission of shock waves which are impacting on a water-gas interface from gas or water medium.

• Journal of the Korean Society of Safety
• /
• v.23 no.3
• /
• pp.87-92
• /
• 2008

A finite element method which is discontinuous in time is developed for the solution of the classical parabolic model of heat conduction problems. The approximations are continuous with respect to the space variables for each fixed time, but they admit discontinuities with respect to the time variable at each time step. The method is superior to other well-known approaches to these problems in that it allows a wider range of moving boundary value problems to be dealt with, such as are encountered in complex engineering operations like ground freezing. The method is applied to one-dimensional and two-dimensional heat conduction problems in this paper, although it could be extended to more higher dimensional problems. Several example problems are discussed and illustrated, and comparisons are made with analytical approaches where these can also be used.

• Journal of computational fluids engineering
• /
• v.4 no.1
• /
• pp.19-26
• /
• 1999

This paper describes a numerical method for predicting the incompressible unsteady laminar three-dimensional flows with free-surface. The Navier-Stokes equations governing the flows have been discretized by means of finite-difference approximations, and the resulting equations have been solved via the SIMPLE-C algorithm. The free-surface is defined by the motion of a set of marker particles and the interface behaviour was investigated by means of a "Lagrangian" technique. Using the GALA concept of Spalding, the conventional mass continuity equation is modified to form a volumetric or bulk-continuity equation. The use of this bulk-continuity relation allows the hydrodynamic variables to be computed over the entire flow domain including both liquid and gas regions. Thus, the free-surface boundary conditions are imposed implicitly and the problem formulation is greatly simplified. The numerical procedure is validated by comparing the predicted results of a periodic standing waves problems with analytic solutions. The results show that this numerical method produces accurate and physically realistic predictions of three-dimensional free-surface flows.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.6 no.1
• /
• pp.12-22
• /
• 1994

Time evolution of linear water waves on a constant depth generated by a disturbance is analyzed by asymptotic methods; stationary phase, steepest descents and leading wave approximation. In order to verify the derived formulae of surface displacements for 1-D and 2-D waves. surface displacements are calculated and plotted from both the formulae and a numerical integration. The existing results for surface displacements are verified in which the leading amplitude of 1-D waves during the evolution decays as f- T/B, the rest of the wavetrain as t$^{-1}$ 2/ and the rest of the wavetrain of 2-D waves as t-1. But it is shown that the leading amplitude of 2-D waves decays as t 5/6 which is different from Kajiura's result t$^{-4}$ 3/.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.23 no.6 s.165
• /
• pp.968-978
• /
• 1999

The effective elastic properties of materials containing spherical inclusions were calculated by the elastic wave scattering theory. In the formulation additional scattering fields by the presence of random multiple scatterers that affects the effective properties were found by the single scattering approximation. In calculating the scattering fields the ensemble average on the displacements and strains inside the scatterer was found from the static approximation at long wavelength limit. The displacements were assumed to be equal to the incident field, while the strains were calculated by Eshelby's equivalent inclusion principle on the single inclusion problem. Four different models were considered and they reflected different degrees of multiple scattering effects based on the approximation introduced in the process of embedding the inclusion in the matrix. The expressions for the effective elastic constants were given in each model, and their relations to the results obtained from other scattering theory and elasticity theory were discussed. The theoretical predictions were compared with experimental results on the epoxy matrix composites containing tungsten particles of different sizes and volume fractions

• Journal for History of Mathematics
• /
• v.25 no.3
• /
• pp.1-14
• /
• 2012

Since JiuZhang SuanShu(九章算術), the basic field of the traditional mathemtics in Eastern Asia is the field of rational numbers and hence irrational solutions of equations should be replaced by rational approximations. Thus approximate solutions of equations became a very important subject in theory of equations. We first investigate the history of approximate solutions in Chinese sources and then compare them with those in Chosun mathematics. The theory of approximate solutions in Chosun has been established in SanHakWonBon(算學原本) written by Park Yul(1621 - 1668) and JuSeoGwanGyun(籌書管見, 1718) by Cho Tae Gu(趙泰耉, 1660-1723). We show that unlike the Chinese counterpart, Park and Cho were concerned with errors of approximate solutions and tried to find better approximate solutions.