• ETRI Journal
• /
• v.15 no.2
• /
• pp.35-51
• /
• 1993

This paper presents function approximation based on nonparametric estimation. As an estimation model of function approximation, a three layered network composed of input, hidden and output layers is considered. The input and output layers have linear activation units while the hidden layer has nonlinear activation units or kernel functions which have the characteristics of bounds and locality. Using this type of network, a many-to-one function is synthesized over the domain of the input space by a number of kernel functions. In this network, we have to estimate the necessary number of kernel functions as well as the parameters associated with kernel functions. For this purpose, a new method of parameter estimation in which linear learning rule is applied between hidden and output layers while nonlinear (piecewise-linear) learning rule is applied between input and hidden layers, is considered. The linear learning rule updates the output weights between hidden and output layers based on the Linear Minimization of Mean Square Error (LMMSE) sense in the space of kernel functions while the nonlinear learning rule updates the parameters of kernel functions based on the gradient of the actual output of network with respect to the parameters (especially, the shape) of kernel functions. This approach of parameter adaptation provides near optimal values of the parameters associated with kernel functions in the sense of minimizing mean square error. As a result, the suggested nonparametric estimation provides an efficient way of function approximation from the view point of the number of kernel functions as well as learning speed.

• Wind and Structures
• /
• v.9 no.6
• /
• pp.441-455
• /
• 2006

A more applicable optimization model for extracting flutter derivatives of bridge decks is presented, which is suitable for time-varying weights for fitting errors and different lengths of vertical bending and torsional free vibration data. A stochastic search technique for searching the optimal solution of optimization problem is developed, which is more convenient in understanding and programming than the alternate iteration technique, and testified to be a valid and efficient method using two numerical examples. On the basis of the section model test of Sutong Bridge deck, the flutter derivatives are extracted by the stochastic search technique, and compared with the identification results using the modified least-square method. The Empirical Mode Decomposition method is employed to eliminate noise, trends and zero excursion of the collected free vibration data of vertical bending and torsional motion, by which the identification precision of flutter derivatives is improved.

• KSBB Journal
• /
• v.7 no.2
• /
• pp.112-117
• /
• 1992

An optimal composition of medium for hyaluronic acid production and some characteristics of its from Streptococcus zooepidemicus were investigated. The hyaluronic acid from S. zooepidemicus was reached maximum level in the BY-medium containing 0.1% beef extract, 0.1% yeast extract, 3.0% glucose, 2.0% peptone, 0.1% NaCl and $0.5%CaCO_3$ (pH 7.5) at $37^{\circ}C$ for 36 hours with shaking. Addition of $CaCO_3$ to the medium was necessary to neulralize the lowered pH which was resulted from hyaluronic acid production. Molecular weights of extracelluar and cellular hyaluronic acid produced by the strain were $1-1.4{\times }10^6$ and $5{\times}10^6$, respectively. The amount of extracellular hyaluronic acid was 91.9% of total hyaluronic acid produced and the vest was all intracellular.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.25 no.1
• /
• pp.97-104
• /
• 2015

An 'associative cube', a class of auto-associative memories, is revisited here, in which training data and hidden orthogonal basis functions such as wavelet packets or Fourier kernels, are combined in the weight cube. This weight cube has hidden units in its depth, represented by a three dimensional cubic structure. We develop an unsupervised incremental learning mechanism based upon the adaptive least squares method. Training data are mapped into orthogonal basis vectors in a least-squares sense by updating the weights which minimize an energy function. Therefore, a prescribed orthogonal kernel is incrementally assigned to an incoming data. Next, we show how a decoding procedure finds the closest one with a competitive network in the hidden layer. As noisy test data are applied to an associative cube, the nearest one among the original training data are restored in an optimal sense. The simulation results confirm robustness of associative cubes even if test data are heavily distorted by various types of noise.

• KSBB Journal
• /
• v.22 no.1
• /
• pp.53-57
• /
• 2007

We have studied biochemical characterization of lectin purified from kidney bean seedling through PBS extraction, $(NH_4)_2SO_4$ precipitation, and Sephadex G-100 affinity chromatography. The lectin was agglutinated by rabbit erythrocytes. This lectin analyzed by SDS-PAGE is a tetramer composed of two subunits with molecular weights of 46 and 44 kDa. The optimal temperature and thermal stability of the lectin was 30$^{\circ}C$ and 40-80$^{\circ}C$, respectively. The maximal pH of this lectin was pH 8.2.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.33 no.4
• /
• pp.69-78
• /
• 2010

Multi-criteria ABC inventory classification is one of the most widely employed techniques for efficient inventory control, and it considers more than one criterion for categorizing inventory items into groups of different importance. Recently, Ramanathan (2006) proposed a weighted linear optimization (WLO) model for the problem of multi-criteria ABC inventory classification. The WLO model generates a set of criteria weights for each item and assigns a normalized score to each item for ABC analysis. Although the WLO model is considered to have many advantages, it has a limitation that many items can share the same optimal efficiency score. This limitation can hinder a precise classification of inventory items. To overcome this deficiency, we propose a context-dependent DEA based method for multi-criteria ABC inventory classification problems. In the proposed model, items are first stratified into several efficiency levels, and then the relative attractiveness of each item is measured with respect to less efficient ones. Based on this attractiveness measure, items can be further discriminated in terms of their importance. By a comparative study between the proposed model and the WLO model, we argue that the proposed model can provide a more reasonable and accurate classification of inventory items.

• Journal of Korea Multimedia Society
• /
• v.2 no.2
• /
• pp.129-136
• /
• 1999

A linear feature often provide sufficient information for image understanding and coding. An objective of the research reported in this paper is to develop and analyze the reliable methods of extracting lines in gray scale images. The Hough Transform is known as one of the optimal paradigms to detect or identify the linear features by transforming edges in images into peaks in parameter space. The scheme proposed here uses the fuzzy gradient direction model and weights the gradient magnitudes for deciding the voting values to be accumulated in parameter space. This leads to significant computational savings by restricting the transform to within some support region of the observed gradient direction which can be considered as a fuzzy variable and produces robust results.

• Proceedings of the KIEE Conference
• /
• 2003.11b
• /
• pp.171-174
• /
• 2003

In this paper, a Learning module for a reconfigurable neural network processor(ERNIE) was proposed for an On-chip learning. The existing reconfigurable neural network processor(ERNIE) has a much better performance than the software program but it doesn't support On-chip learning function. A learning module which is based on Back Propagation algorithm was designed for a help of this weak point. A pipeline structure let the learning module be able to update the weights rapidly and continuously. It was tested with five types of alphabet font to evaluate learning module. It compared with C programed neural network model on PC in calculation speed and correctness of recognition. As a result of this experiment, it can be found that the neural network processor(ERNIE) with learning module decrease the neural network training time efficiently at the same recognition rate compared with software computing based neural network model. This On-chip learning module showed that the reconfigurable neural network processor(ERNIE) could be a evolvable neural network processor which can fine the optimal configuration of network by itself.

• Proceedings of the KSME Conference
• /
• 2007.05a
• /
• pp.96-101
• /
• 2007

A new metallic sandwich panel with a quasi-Kagome truss core subjected to bending load has been analyzed. First, equations of the failure loads corresponding to the eight failure modes are presented. Then, non-dimensional forms of the equations are derived as functions of three geometric variables, one material parameter (yield strain), one load index and one weight index. Failure maps are presented for a given weight index. By using the dimensionless forms of equations as the design constraints, two kinds of optimization are performed. One is based on the weight, that is, the objective function, namely, the dimensionless load is to be maximized for a given weight. Another is based on the load, that is, the dimensionless weight is to be minimized for a given load. The results of the two optimization processes are found to agree each other. The optimized geometric variables are derived as a function of given weights or failure loads. The performance of the quasi-Kagome truss as the core of a sandwich panel is evaluated by comparison with those of honeycomb cored and octet truss cored panels

• Smart Structures and Systems
• /
• v.21 no.1
• /
• pp.123-137
• /
• 2018

Structural monitoring is the most important part of the construction and operation of the embankment dams. Appropriate instruments selection for dams is vital, as inappropriate selection causes irreparable loss in critical condition. Due to the lack of a systematic approach to determine adequate instruments, a framework based on three comparable Multi-Attribute Decision Making (MADM) methods, which are VIKOR, technique of order preference by similarity to ideal solution (TOPSIS) and Preference ranking organization method for enrichment evaluation (PROMETHEE), has been developed. MADM techniques have been widely used for optimizing priorities and determination of the most suitable alternatives. However, the results of the different methods of MADM have indicated inconsistency in ranking alternatives due to closeness of judgements from decision makers. In this study, 9 criteria and 42 geotechnical instruments have been applied. A new method has been developed to determine the decision makers' importance weights and an aggregation method has been introduced to optimally select the most suitable instruments. Consequently, the outcomes of the aggregation ranking correlate about 94% with TOPSIS and VIKOR, and 83% with PROMETHEE methods' results providing remarkably appropriate prioritisation of instruments for embankment dams.