• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.8
• /
• pp.1891-1898
• /
• 2013

To efficiently execute 3D graphic APIs, such as OpenGL and Direct3D, special purpose arithmetic unit(SFU) which supports floating-point sine, cosine, reciprocal, inverse square root, base-two exponential, and logarithmic operations is designed. The SFU uses second order minimax approximation method and lookup table method to satisfy both error less than 2 ulp(unit in the last place) and high speed operation. The designed circuit has about 2.3-ns delay time under 65nm CMOS standard cell library and consists of about 23,300 gates. Due to its maximum performance of 400 MFLOPS and high accuracy, it can be efficiently applicable to mobile 3D graphics application.

• Journal of the Korean Data and Information Science Society
• /
• v.17 no.2
• /
• pp.543-557
• /
• 2006

This paper is concerned with the applicability of the chi-square approximation to the six disparity statistics: the Pearson chi-square, the generalized likelihood ratio, the power divergence, the blended weight chi-square, the blended weight Hellinger distance, and the negative exponential disparity statistic. Three dimensional contingency tables of small and moderate sample sizes are generated to be fitted to all possible hierarchical log-linear models: the completely independent model, the conditionally independent model, the partial association models, and the model with one variable independent of the other two. For models with direct solutions of expected cell counts, point estimates and confidence intervals of the 90 and 95 percentage points of six statistics are explored. For model without direct solutions, the empirical significant levels and the empirical powers of six statistics to test the significance of the three factor interaction are computed and compared.

• Journal of applied mathematics & informatics
• /
• v.10 no.1_2
• /
• pp.245-250
• /
• 2002

Consider the Convex Polygon Pm={Al , A2, ‥‥, Am} With Vertex points A$\_$i/ = (a$\_$i/, b$\_$i/),i : 1,‥‥, m, interior P$\^$0/$\_$m/, and length of perimeter denoted by L(P$\_$m/). Let R$\_$n/ = {B$_1$,B$_2$,‥‥,B$\_$n/), where B$\_$i/=(x$\_$i/,y$\_$I/), i =1,‥‥, n, denote a regular polygon with n sides of equal length and equal interior angle. Kaiser[4] used the regular polygon R$\_$n/ to approximate P$\_$m/, and the problem examined in his work is to position R$\_$n/ with respect to P$\_$m/ to minimize the area of the symmetric difference between the two figures. In this paper we give the quality of a approximating regular polygon R$\_$n/ to approximate P$\_$m/.

• ETRI Journal
• /
• v.26 no.4
• /
• pp.297-306
• /
• 2004

A plasma is a collection of charged particles and on average is electrically neutral. In fabricating integrated circuits, plasma etching is a key means to transfer a photoresist pattern into an underlayer material. To construct a predictive model of plasma-etching processes, a polynomial neural network (PNN) is applied. This process was characterized by a full factorial experiment, and two attributes modeled are its etch rate and DC bias. According to the number of input variables and type of polynomials to each node, the prediction performance of the PNN was optimized. The various performances of the PNN in diverse environments were compared to three types of statistical regression models and the adaptive network fuzzy inference system (ANFIS). As the demonstrated high-prediction ability in the simulation results shows, the PNN is efficient and much more accurate from the point of view of approximation and prediction abilities.

• Journal of computational fluids engineering
• /
• v.11 no.1 s.32
• /
• pp.21-28
• /
• 2006

Reliability Based Design Optimization(RBDO) is one of the optimization methods that minimize the product failure due to small changes of operating conditions or process errors. It searches the optimum that satisfies the safety margin of each constraint, and it gives stable and reliable designs. However, RBDO requires many times oj computational efforts compared with the conventional deterministic optimization(DO) to evaluate the probability of failure about each constraint, therefore it is hard to apply directly to large-scaled problems such as a flexible wing shape design optimization. For the efficient reliability analysis, the approximate reliability analysis method with the two-point approximation(TPA) is proposed In this study, the lift-to-drag ratio maximization designs are performed with 3-dimensional Navier-Stokes analysis and NASTRAN structural analysis, and the optimization results about the deterministic, FORM and SORM are compared.

• International Journal of Aeronautical and Space Sciences
• /
• v.16 no.4
• /
• pp.510-519
• /
• 2015

Concerning the issue of high-dimensions, hybrid uncertainties of randomness and intervals including implicit and highly nonlinear limit state function, reliability analysis based on the hybrid uncertainty reliability mode combining with back propagation neural network (HU-BP neural network) is proposed in this paper. Random variables and interval variables are as input layer of the neural network, after the training and approximation of the neural network, the response variables are obtained through the output layer. Reliability index is calculated by solving the optimization model of the most probable point (MPP) searching in the limit state band. Two numerical cases are used to demonstrate the method proposed in this paper, and finally the method is employed to solving an engineering problem of the aerospace friction plate. For this high nonlinear, small failure probability problem with interval variables, this method could achieve a good analysis result.

• Journal of Korea Technology Innovation Society
• /
• v.8 no.3
• /
• pp.887-910
• /
• 2005

This paper attempts to approximate the distribution function for the number of innovation activities (NIA). To this end, the dataset of 2002 Korean Innovation Survey (KIS 2002) published by Science and Technology Policy Institute is used. To deal with zero NTI values given by a considerable number of firms in the KIS 2002 survey, a mixture model of distributions for NIA is applied. The NIA is specified as a mixture of two distributions, one with a point mass at zero and the other with full support on the positive half of the real line. The model was empirically verified for the KIS 2002 data. The mixture model can easily capture the common bimodality feature of the NIA distribution. In addition, when covariates were added to the mixture model, it was found that the probability that a firm has zero NIA significantly varies with some variables.

• Proceedings of the KIEE Conference
• /
• 1998.11b
• /
• pp.603-605
• /
• 1998

In recent years, chip delay estimation has had an increasingly important impact on overall design technology. The analysis of the timing behavior of an ASIC should be based not only on the delay characteristics of gates and interconnect circuits but also on the interactions between them. This model plays an important role in our CAD system to analyze the ASIC timing characteristics accurately, together with two-dimensional gate delay table model, AWE algorithm and effective capacitance concept. In this paper, we present an efficient algorithm which accounts for series resistance by computing a reduced-order approximation for the driving-point admittance of an RC-tree and an effective capacitance equation that captures the complete waveform response accurately.

• Tunnel and Underground Space
• /
• v.25 no.1
• /
• pp.56-67
• /
• 2015

In this study, a source locating technique applicable to transversely isotropic media was developed. Wave velocity anisotropy was considered based on the partition approximation method, which simply enabled AE source locating. Sets of P wave arrival time were decided by the two-step AIC algorithm and they were later used to locate the AE sources when having the least error compared with the partitioned elements. In order to validate the technique, pencil lead break test on artificial transversely isotropic mortar specimen was carried out. Defining the absolute error as the distance between the pencil lead break point and the located point, 1.60 mm ~ 14.46 mm of range and 8.57 mm of average were estimated therefore it was regarded as thought to be 'acceptable' considering the size of the specimen and the AE sensors. Comparing each absolute error under different threshold levels, results showed small discrepancies therefore this technique was hardly affected by background noise. Absolute error could be decomposed into each coordinate axis error and through it, effect of AE sensor position could be understood so if optimum sensor position was able to be decided, one could get more precise outcome.

• Journal of the Korean Data and Information Science Society
• /
• v.9 no.2
• /
• pp.345-355
• /
• 1998

The back propagation(BP) algorithm allows multilayer feedforward neural networks to learn input-output mappings from training samples. It iteratively adjusts the network parameters(weights) to minimize the sum of squared approximation errors using a gradient descent technique. However, the mapping acquired through the BP algorithm may be corrupt when errorneous training data are employed. In this paper two types of robust backpropagation algorithms are discussed both from a theoretical point of view and in the case studies of nonlinear regression function estimation and handwritten Korean character recognition. For future research we suggest Bayesian learning approach to neural networks and compare it with two robust backpropagation algorithms.