• The Korean Journal of Applied Statistics
• /
• v.28 no.2
• /
• pp.281-288
• /
• 2015

In this paper, we study an approximate procedure to evaluate a penalized maximum likelihood estimator (MLE) for a mixed effects model. The procedure approximates the Hessian matrix of the penalized MLE with a structured sparse matrix or an arrowhead type matrix to speed its computation. In this paper, we numerically investigate the gain in computation time as well as approximation error from the considered approximation procedure.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2012.05a
• /
• pp.710-710
• /
• 2012

구형펄스모형(Rectangular Pulse Model)에서 반영하지 못하는 강우의 군집특성을 잘 반영하는 NSRPM(Neyman-Scott Rectangular Pulse Model) 강우생성 모형은 수자원 분야에 널리 쓰이고 있다. 일반적으로 NSRPM의 5개의 매개변수를 추정하는 최적화기법으로 DFP(Davidon-Fletcher-Powell)과 유전자알고리즘(Genetic Algorithm)을 사용하고 있다. 그러나 DFP는 주어진 초기 값에 따라 민감하며 각 반복 단계마다 헤시안행렬(Hessian Matrix)을 계산하여야 하며 추정된 전체의 해가 국지해에 수렴 할 수 있는 단점이 있다. 유전자 알고리즘을 DFP와 다르게 헤시안 행렬을 사용하지 않고 최적화를 할 수 있다는 장점이 있으나 시간이 오래 걸리는 단점이 있다. 이에 본 연구에서는 이러한 단점을 보완, 강화 하기위해서 최적화 기법으로 반복 단계마다 미분계산이 필요하지 않고 빠른 속도로 계산이 가능한 Nelder-Mead 알고리즘 이용하여 NSRPM매개변수를 추정하고 정확도를 비교하였다. 표 1은 각 기법을 이용하여 추정된 매개변수를 이용하여 생성한 강우의 통계특성과 관측된 통계특성의 상대오차를 나타낸 것이다. 괄호 안 숫자는 중첩되지 않는 누적시간을 나타낸다. 상대오차는 다음과 같다(식 1). 분석결과 Nelder-Mead 기법이 1시간의 평균, 공분산과 6시간 분산 등 전체적으로 GA, DFP보다 높은 정확도를 보였다.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.20 no.12
• /
• pp.3804-3821
• /
• 1996

The configuration optimization is a structural optimization method which includes the coordinates of a structure as well as the sectional properties in the design variable set. Effective reduction of the weight of discrete structures can be obrained by changing the geometry while satisfying stress, Ei;er bickling, displacement, and frequency constraints, etc. However, the nonlinearity due to the configuration variables may cause the difficulties of the convergence and expensive computational cost. An efficient approximation method for the configuration optimization has been developed to overcome the difficulties. The method approximates the constraint functions based onthe second-order Taylor series expansion with reciprocal design variables. The Hessian matrix is approzimated from the information on previous design points. The developed algotithms are coded and the examples are solved.

• Bulletin of the Korean Mathematical Society
• /
• v.46 no.1
• /
• pp.35-43
• /
• 2009

We say that an isometric immersed hypersurface x : $M^n\;{\rightarrow}\;{\mathbb{R}}^{n+1}$ is of $L_k$-finite type ($L_k$-f.t.) if $x\;=\;{\sum}^p_{i=0}x_i$ for some positive integer p < $\infty$, $x_i$ : $M{\rightarrow}{\mathbb{R}}^{n+1}$ is smooth and $L_kx_i={\lambda}_ix_i$, ${\lambda}_i\;{\in}\;{\mathbb{R}}$, $0{\leq}i{\leq}p$, $L_kf=trP_k\;{\circ}\;{\nabla}^2f$ for $f\;{\in}\'C^{\infty}(M)$, where $P_k$ is the kth Newton transformation, ${\nabla}^2f$ is the Hessian of f, $L_kx\;=\;(L_kx^1,\;{\ldots},\;L_kx^{n+1})$, $x=(x^1,\;{\ldots},\;x^{n+1})$. In this article we study the following(hyper)surfaces in ${\mathbb{R}}^{n+1}$ from the view point of $L_1$-finiteness type: totally umbilic ones, generalized cylinders $S^m(r){\times}{\mathbb{R}}^{n-m}$, ruled surfaces in ${\mathbb{R}}^{n+1}$ and some revolution surfaces in ${\mathbb{R}}^3$.

• Journal of Drive and Control
• /
• v.14 no.4
• /
• pp.45-50
• /
• 2017

The gas spring is a hydropneumatic adjusting element, consisting of a pressure tube, a piston rod, a piston and a connection fitting. The gas spring is filled with compressed nitrogen within the cylinder. The filling pressure acts on both sides of the piston and because of area difference it produces an extension force. Therefore, a gas spring is similar in function compare to mechanical coil spring. Conversely, optimization is a process of finding the best set of parameters to reach a goal while not violating certain constraints. The AMESim software provides NLPQL (Nonlinear Programming by Quadratic Lagrangian) and GA (genetic algorithm) for optimization. The NLPQL method builds a quadratic approximation to the Lagrange function and linear approximations to all output constraints at each iteration, starting with the identity matrix for the Hessian of the Lagrangian, and gradually updating it using the BFGS method. On each iteration, a quadratic programming problem is solved to find an improved design until the final convergence to the optimum design. In this study, we conducted optimization design of the gas spring reaction force with NLPQL.

• The Korean Journal of Applied Statistics
• /
• v.21 no.2
• /
• pp.327-340
• /
• 2008

This paper derives joint and conditional Lagrange multiplier tests based on information matrix for testing functional form and/or the presence of autocorrelation in a regression model. Small sample properties of these tests are assessed by Monte Carlo study and comparisons are made with LM tests based on Hessian matrix. The results show that the proposed $LM_E$ tests have the most appropriate finite sample performance.

• Journal of Biosystems Engineering
• /
• v.40 no.3
• /
• pp.212-223
• /
• 2015

Purpose: Physical properties of Detarium microcarpum seeds were investigated as a function of moisture content to explore the possibility of developing bulk handling and processing equipment. Methods: Seed size, surface area, and 1,000-seed weight were determined by measuring the three principal axes, measuring area on a graph paper, and counting and weighing seeds. Particle and bulk densities were determined using liquid displacement and weight in a measuring cylinder, respectively. Porosity was computed from particle and bulk densities. Roundness and sphericity were measured using shadowgraphs. Angle of repose and static and kinetic coefficients of friction were determined using the vertical cylindrical pipe method, an inclined plane, and a kinetic coefficient of friction apparatus. Results: In the moisture range of 8.2%-28.5% (db), the major, intermediate, and the minor axes increased from 2.95 to 3.21 cm, 1.85 to 2.61 cm, and 0.40 to 1.21 cm, respectively. Surface area, 1,000-seed weight, particle density, porosity, and angle of repose increased from 354.62 to $433.19cm^2$, 3.184 to 3.737 kg, 1060 to $1316kg/m^3$, and 30.0% to 53.1%, respectively, whereas bulk density decreased from 647.6 to $617.2kg/m^3$. Angle of repose increased from $13.9^{\circ}$ to $28.4^{\circ}$. Static and kinetic coefficients of friction varied between 0.096 and 0.638 on different structural surfaces. Conclusions: Arithmetic mean, geometric mean, and equivalent sphere effective diameters determined at the same moisture level were significantly different from each other, with the arithmetic mean diameter being greatest. Surface area, 1,000-seed weight, particle density, porosity, and angle of repose all increased linearly with moisture content. Bulk density decreased linearly with moisture content. The coefficients of friction had linear relationships with moisture content. The highest values of static and kinetic coefficients of friction were observed on galvanized steel and hessian fabric, respectively, whereas the lowest values were observed on fiberglass.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.4
• /
• pp.634-640
• /
• 2018

In this paper, we propose an automatic extraction algorithm of region of interest(ROI) based on medical x-ray images. The proposed algorithm uses segmentation, feature extraction, and reference image matching to detect lesion sites in the input image. The extracted region is searched for matching lesion images in the reference DB, and the matched results are automatically extracted using the Kalman filter based fitness feedback. The proposed algorithm is extracts the contour of the left hand image for extract growth plate based on the left x-ray input image. It creates a candidate region using multi scale Hessian-matrix based sessionization. As a result, the proposed algorithm was able to split rapidly in 0.02 seconds during the ROI segmentation phase, also when extracting ROI based on segmented image 0.53, the reinforcement phase was able to perform very accurate image segmentation in 0.49 seconds.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.4
• /
• pp.501-507
• /
• 2021

Neural networks with differentiable activation functions are differentiable with respect to input variables. We improve the approximation capability of neural networks by using the gradient and Hessian of neural networks to satisfy the differential equations of the problems of interest. We apply differential neural networks to the pricing of financial options, where stochastic differential equations and the Black-Scholes partial differential equation represent the differential relation of price of option and underlying assets, and the first and second derivatives of option price play an important role in financial engineering. The proposed neural network learns - (a) the sample paths of option prices generated by stochastic differential equations and (b) the Black-Scholes equation at each time and asset price. Experimental results show that the proposed method gives accurate option values and the first and second derivatives.

• Journal of Internet of Things and Convergence
• /
• v.10 no.3
• /
• pp.25-33
• /
• 2024

In tomographic image reconstruction, the focus is on developing CT image reconstruction methods that can maintain high image quality while reducing patient radiation exposure. Typically, statistical image reconstruction methods have the ability to generate high-quality and accurate images while significantly reducing patient radiation exposure. However, in cases like CT image reconstruction, which involve multi-dimensional parameter estimation, the degree of the Hessian matrix of the penalty function is very large, making it impossible to calculate. To solve this problem, the author proposed the PEMG-1 algorithm. However, the PEMG-1 algorithm has issues with the convergence speed, which is typical of statistical image reconstruction methods, and increasing the penalty log-likelihood. In this study, we propose a reconstruction algorithm that ensures fast convergence speed and monotonic increase in likelihood. The basic structure of this algorithm involves sequentially updating groups of pixels instead of updating all parameters simultaneously with each iteration.