• Proceedings of the Korean Statistical Society Conference
• /
• 2000.11a
• /
• pp.135-140
• /
• 2000

베이지안 신경망 모형(Bayesian Neural Networks Models)에서 주어진 입력값(input)은 블랙 박스(Black-Box)와 같은 신경망 구조의 각 층(layer)을 거쳐서 출력값(output)으로 계산된다. 새로운 입력 데이터에 대한 예측값은 사후분포(posterior distribution)의 기대값(mean)에 의해 계산된다. 주어진 사전분포(prior distribution)와 학습데이터에 의한 가능도함수(likelihood functions)를 통해 계산되어진 사후분포는 매우 복잡한 구조를 갖게 됨으로서 기대값의 적분계산에 대한 어려움이 발생한다. 이때 확률적 추정에 의한 근사 방법인 몬테칼로 적분을 이용한다. 이러한 방법으로서 Hybrid Monte Carlo 알고리즘은 우수한 결과를 제공하여준다(Neal 1996). 본 논문에서는 Hybrid Monte Carlo 알고리즘과 기존에 많이 사용되고 있는 Gibbs sampling, Metropolis algorithm, 그리고 Slice Sampling등의 몬테칼로 방법들을 비교한다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1996.06a
• /
• pp.69-75
• /
• 1996

40-100m 수심에 평탄한 모래층으로 구성된 구성된 친해를 1/1,000모형으로 축소시킨 무향수조내에서 Pekeris 모델에 대한 normal mode 관측실험을 점수신기로 수행하였다. 무향수조에 사용된 쐐기형 흡음재가 50kHz이상에서 20-25dB의 반향음 감쇠율을 가지므로 해서 이 수조가 해양과 유사한 조건을 갖게 되었으며, 실험에 사용된 점수신기는 B&K의 8130 수중청음기에 음향중심으로부터 조그만 크기의 사각형 틈을 제외하고 pressure-release 물질을 씌움으로써 얻을 수 있었다. 이 실험을 통해 얻은 결과들을 종합해 보면 연속모드의 피적분함수가 cut-off로부터 멀어지면 불연속모드의 합으로만 된 이론식과는 대략 수심의 10배 이내에서 다소 차이를 보일 뿐 전 관측거리에서 실험결과와 비교적 잘 일치하고 있다. 그러나 연속모드의 공진이 cut-off에 가까워 질수록 가상모드가 음향에 미치는 역할이 증대되어 근거리에서는 불연속모드의 합으로만 된 이론식과 큰 차이가 날 뿐만 아니라 영향이 미치는 거리가 수심의 수십배까지 증가함을 볼 수 있다. 특히 불연속모드의 개수가 2개이고 cut-off로부터 아주 가까운 가상모드가 존재할 경우, 가상모드가 수심의 60배 가까운 거리까지 영향을 미치는 것을 확인할 수 있다.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.11 no.3
• /
• pp.149-155
• /
• 1999

A treatment of wave breaking is included in the extended Boussinesq equations of Nwogu. A spatially distributed source function and sponge layers are used to reduce the reflected waves in the computa¬tional domain. The model uses fourth-order Adams predictor-corrector method to advance in time, and discretizes first-order spatial derivatives to fourth-order accuracy, and thus reducing all truncation errors to a level smaller than the dispersive terms. The generated wave fields are found to be good and the corresponding wave heights are very close to their target values. For the tests of wave breaking, although agreement is considered to be reasonable, wave heights in the inner surf zone are over-predicted. This indicates the breaking parameters in the model should be adjusted.

• Journal of Korea Water Resources Association
• /
• v.50 no.3
• /
• pp.201-210
• /
• 2017

This study proposes a numerical model which is able to simulate turbidity currents intruding into a reservoir and resulting sediment depositions. The proposed model is applied to laboratory experiments by Toniolo and Schultz (2005), and propagation of turbidity currents, morphological change, and trap of suspended sediment are simulated. It is simulated that the turbidity current after plunging at the foreset of the model delta, propagates along the bottom. The thickness of the turbidity current increases significantly after being blocked by the dam, and this effect is propagated in the upstream direction. In addition, it is simulated that the foreset moves in the downstream direction due to both the bedload and suspended load and the thickness of the bottom set increases due to the suspended load. It is found that the height of the intake affects the thickness of the turbidity current and the location of the internal hydraulic jump. The impact of the height of the intake on the trap efficiency is not clear in the experimental results, however, overall trap efficiency is predicted quite successfully by the model. Also, sensitivity analysis is carried out, and the results indicates that the particle size affects the trap efficiency most.

• Journal of the Korean earth science society
• /
• v.43 no.6
• /
• pp.691-711
• /
• 2022

The state-of-the-art Earth system model as a virtual Earth is required for studies of current and future climate change or climate crises. This complex numerical model can account for almost all human activities and natural phenomena affecting the atmosphere of Earth. The Unified Model (UM) from the United Kingdom Meteorological Office (UK Met Office) is among the best Earth system models as a scientific tool for studying the atmosphere. However, owing to the expansive numerical integration cost and substantial output size required to maintain the UM, individual research groups have had to rely only on supercomputers. The limitations of computer resources, especially the computer environment being blocked from outside network connections, reduce the efficiency and effectiveness of conducting research using the model, as well as improving the component codes. Therefore, this study has presented detailed guidance for installing a new version of the UM on high-performance parallel computers (Linux clusters) owned by individual researchers, which would help researchers to easily work with the UM. The numerical integration performance of the UM on Linux clusters was also evaluated for two different model resolutions, namely N96L85 (1.875° ×1.25° with 85 vertical levels up to 85 km) and N48L70 (3.75° ×2.5° with 70 vertical levels up to 80 km). The one-month integration times using 256 cores for the AMIP and CMIP simulations of N96L85 resolution were 169 and 205 min, respectively. The one-month integration time for an N48L70 AMIP run using 252 cores was 33 min. Simulated results on 2-m surface temperature and precipitation intensity were compared with ERA5 re-analysis data. The spatial distributions of the simulated results were qualitatively compared to those of ERA5 in terms of spatial distribution, despite the quantitative differences caused by different resolutions and atmosphere-ocean coupling. In conclusion, this study has confirmed that UM can be successfully installed and used in high-performance Linux clusters.

• The KIPS Transactions:PartB
• /
• v.8B no.3
• /
• pp.231-236
• /
• 2001

일반적인 다층 신경망에서 가중치의 갱신 알고리즘으로 사용하는 오류 역전과 방식은 가중치 갱신 결과를 고정된(fixed) 한 개의 값으로 결정한다. 이는 여러 갱신의 가능성을 오직 한 개의 값으로 고정하기 때문에 다양한 가능성들을 모두 수용하지 못하는 면이 있다. 하지만 모든 가능성을 확률적 분포로 표현하는 갱신 알고리즘을 도입하면 이런 문제는 해결된다. 이러한 알고리즘을 사용한 베이지안 신경망 모형(Bayesian Neural Networks Models)은 주어진 입력값(Input)에 대해 블랙 박스(Black-Box)와같은 신경망 구조의 각 층(Layer)을 거친 출력값(Out put)을 계산한다. 이 때 주어진 입력 데이터에 대한 결과의 예측값은 사후분포(posterior distribution)의 기댓값(mean)에 의해 계산할 수 있다. 주어진 사전분포(prior distribution)와 학습데이터에 의한 우도함수(likelihood functions)에 의해 계산한 사후확률의 함수는 매우 복잡한 구조를 가짐으로 기댓값의 적분계산에 대한 어려움이 발생한다. 따라서 수치해석적인 방법보다는 확률적 추정에 의한 근사 방법인 몬테 칼로 시뮬레이션을 이용할 수 있다. 이러한 방법으로서 Hybrid Monte Carlo 알고리즘은 좋은 결과를 제공하여준다(Neal 1996). 본 논문에서는 Hybrid Monte Carlo 알고리즘을 적용한 신경망이 기존의 CHAID, CART 그리고 QUEST와 같은 여러 가지 분류 알고리즘에 비해서 우수한 결과를 제공하는 것을 나타내고 있다.

• Journal of Korea Water Resources Association
• /
• v.48 no.11
• /
• pp.891-903
• /
• 2015

This paper presents numerical simulations of flow around a circular patch of vegetation using a depth-averaged two-dimensional numerical model which is capable of simulating flow structure in vegetated open channel. In order to account for vegetation effect, drag force terms are included in governing equations. Numerical simulations are conducted with various solid volume fractions (SVF). Flow passes through a circular patch and low velocity region, which is called wake region, is formed downstream of the patch. When SVF is larger than 0.08, a recirculation is observed. The location of the recirculation is moved further downstream as SVF decreases. Von-$K{\acute{a}}rm{\acute{a}}n$ vortex street is developed beyond the wake region due to interaction between two shear layers induced by a circular patch of vegetation. The vortex is developed as SVF is larger than 0.08, and the location of the vortex is consistent with the maximum of turbulence kinetic energy. The location of the peak of turbulence kinetic energy is moved further downstream as SVF decreases.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.19 no.1
• /
• pp.81-96
• /
• 2007

Nonlinear shoaling characteristics over surf zone are numerically investigated based on spatially averaged NavierStokes equation. We also test the validity of gradient model for turbulent stresses due to wave breaking using the data acquainted during SUPERTANK LABORATORY DATA COLLECTION PROJECT(Krauss et al., 1992). It turns out that the characteristics length scale of breaking induced current is not negligible, which firmly stands against ever popular gradient model, ${\kappa}-{\varepsilon}$ model, but favors Large Eddy Simulation with finer grid. Based on these observations, we model the residual stress of spatially averaged NavierStokes equation after Lagrangian Dynamic Smagorinsky(Meneveau et al., 1996). We numerically integrate newly proposed wave equations using SPH with Gaussian kernel function. Severely deformed water surface profile, free falling water particle, queuing splash after landing of water particle on the free surface and wave finger due to structured vortex on rear side of wave crest(Narayanaswamy and Dalrymple, 2002) are successfully duplicated in the numerical simulation of wave propagation over uniform slope beach, which so far have been regarded very difficult features to mimic in the computational fluid mechanics.

• Geophysics and Geophysical Exploration
• /
• v.10 no.2
• /
• pp.138-146
• /
• 2007

This paper describes 2.5D induced polarization (IP) modeling and inversion algorithms using complex resistivity. The complex resistivity method has merits for acquiring more valuable information about hydraulic parameters and pore fluid than the conventional IP methods. The IP modeling and inversion algorithms are developed by allowing complex arithmetic in existing DC modeling and inversion algorithms. The IP modeling and inversion algorithms use a 2.5D DC finite-element algorithm and a damped least-squares method with smoothness constraints, respectively. The accuracy of the IP modeling algorithm is verified by comparing its responses of two synthetic models with two different approaches: linear filtering for a three-layer model and an integral equation method for a 3D model. Results from these methods are well matched to each other. The inversion algorithm is validated by a synthetic example which has two anomalous bodies, one is more conductive but non-polarizable than the background, and the other is polarizable but has the same resistivity as the background. From the inverted section, we can cleary identify each anomalous body with different locations. Furthermore, in order to verify its efficiency to the real filed example, we apply the inversion algorithm to another three-layer model which includes phase anomaly in the second layer.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.1 no.4
• /
• pp.45-58
• /
• 1997

This paper describes a series of shaking table and pseudodynamic tests for evaluation of seismic performance of base-isolated structures subjected to various seismic earthquake inputs. The main objectives of this study are : (1) evaluation of the effectiveness of base-isolation systems for low-rise structures against severe seismic loads through shaking table tests, (2) verification of the substructuring pseudodynamic test method for the base-isolated structures in comparison with the shaking table test results. In the shaking table test, a quarter scaled three-story structure base-isolated by laminated rubber bearings is tested. In the pseudodynamic test, only the laminated rubber bearing s are tested using the substructuring technique, while the concurrent seismic responses of the superstructure are computed using on-line numerical integration. Comparison with the shaking table test results indicates that the substructuring pseudodynamic test method is very effective for determining the dynamic responses of the base-isolated structure.