• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.7
• /
• pp.170-178
• /
• 1994

The nearest neighbor classification rule is widely used because it is not only simple but the error rate is asymptotically less than twice Bayes theoretical minimum error. But the method basically use the whole training patterns as the reference vectors. so that both storage and classification time increase as the number of training patterns increases. LVQ(Learning Vector Quantization) resolved this problem by training the reference vectors instead of just storing the whole training patterns. But it is a heuristic algorithm which has no theoretic background there is no terminating condition and it requires a lot of iterations to get to meaningful result. This paper is to propose a new training method of the reference vectors. which minimize the given error function. The nearest neighbor network,the network version of the nearest neighbor classification rule is proposed. The network is funtionally identical to the nearest neighbor classification rule is proposed. The network is funtionally identical to the nearest neighbor classification rule and the reference vectors are represented by the weights between the nodes. The network is trained to minimize the error function with respect to the weights by the steepest descent method. The learning algorithm is derived and it is shown that the proposed method can adjust more reference vectors than LVQ in each iteration. Experiment showed that the proposed method requires less iterations and the error rate is smaller than that of LVQ2.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.62 no.11
• /
• pp.1578-1583
• /
• 2013

In this paper, stochastic optimization algorithms of PSO (Particle Swarm Optimization) and APSO (Adaptive Particle Swam Optimization) are studied and compared. It is revealed that the APSO provides faster convergence and better search efficiency than the conventional PSO when they are adopted to find the global minimum of a two-dimensional function. The advantages of the APSO comes from the ability to control the inertia weight, and acceleration coefficients. To verify that the APSO is working better than the standard PSO, the design of a 10GHz microstrip patch as one of the elements of a high frequency array antenna is taken as a test-case and shows the optimized result with 5 iterations in the APSO and 28 iterations in th PSO.

• Steel and Composite Structures
• /
• v.24 no.3
• /
• pp.369-382
• /
• 2017

This study investigates the reliability of the performance levels of moment resisting steel frames subjected to lateral loads such as wind and earthquake. The reliability assessment has been performed with respect to three performance levels: serviceability, damageability, and ultimate limit states. A four-story moment resisting frame is used as a typical example. In the reliability assessment the uncertainties in the loadings and in the capacity of the frame have been considered. The wind and earthquake loads are assumed to have lognormal distribution, and the frame resistance is assumed to have a normal distribution. In order to obtain an appropriate limit state function a linear relation between the loading and the deflection is formed. For the reliability analysis an algorithm has been developed for determination of limit state functions and iterations of the first order reliability method (FORM) procedure. By the method presented herein the multivariable analysis of a complicated reliability problem is reduced to an S-R problem. The procedure for iterations has been tested by a known problem for the purpose of avoiding convergence problems. The reliability indices for many cases have been obtained and also the effects of the coefficient of variation of load and resistance have been investigated.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2003.06a
• /
• pp.9-12
• /
• 2003

This paper describes a step-by-step method to minimize design iterations in a process of product design change. In the design process, two components are coupled if a change of a component can require the other components change, and design iterations are generated by the coupling. The design iteration is one of main factors that increase design effort. In this study, three matrices are used to solve the design iteration of automatic transmission lever, Requirement-Engineering matrix, Engineering-Components matrix, and DSM(Design Structure Matrix). Firstly, with the DSM, the product architecture and conceptual design process are proposed from product function analysis. Secondly, with the QFD, the Requirement-Engineering matrix and Engineering-Components matrix present the relationship among customer requirements, engineering issues, and product components. Lastly, the results of the QFD analysis are used in the DSM to solve the component interactions and to provide design

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.8 no.2
• /
• pp.49-55
• /
• 2008

To address the convergence issue of power control algorithms, a number of algorithms have been developed hat shape the dynamics of up-link power control for cellular network. Power algorithms based on fixed point iterations can be accelerated by the use of various methods, one of the simplest being the use of Newton iterations, however, this method has the disadvantage which not only needs derivatives of the cost function but also may be weak to noisy environment. we showed performance of the power control schemes to solve the fixed point problem under static or stationary channel. They proved goof performance to solve the fixed point problem due to their predictor based optimal control and quadratic convergence rate. Here, we apply the proposed power control schemes to the problem of the dynamic channel or to dynamic time varying link gains. The rigorous simulation results demonstrated the validity of our approach.

• Korean Journal of Mathematics
• /
• v.30 no.3
• /
• pp.459-465
• /
• 2022

We exhibit various properties of the weighted Berezin operator T_α and its iteration T^k_α on L^p(𝜏), where α > -1 and 𝜏 is the invariant measure on the complex unit ball B_n. Iterations of T_α on L¹_R(𝜏) the space of radial integrable functions have performed important roles in proving 𝓜-harmonicity of bounded functions with invariant mean value property. We show differences between the case of 1 < p < ∞ and p = 1, ∞ under the infinite iteration of T_α or the infinite summation of iterations, most of which are extensions or related assertions to the propositions of the previous results.

• Journal of Information Processing Systems
• /
• v.6 no.4
• /
• pp.453-480
• /
• 2010

Cryptographic hash functions reduce inputs of arbitrary or very large length to a short string of fixed length. All hash function designs start from a compression function with fixed length inputs. The compression function itself is designed from scratch, or derived from a block cipher or a permutation. The most common procedure to extend the domain of a compression function in order to obtain a hash function is a simple linear iteration; however, some variants use multiple iterations or a tree structure that allows for parallelism. This paper presents a survey of 17 extenders in the literature. It considers the natural question whether these preserve the security properties of the compression function, and more in particular collision resistance, second preimage resistance, preimage resistance and the pseudo-random oracle property.

• 한국정보통신설비학회:학술대회논문집
• /
• 2006.08a
• /
• pp.202-204
• /
• 2006

A new method is presented to design RF filters without the Similarity Transform of their coupling coefficient matrix as circuit parameters which is very tedious due to pivoting and deciding rotation angles needed during the iterations. The transfer function of a filter is directly used for the design and its desired form is derived by the optimized rational-function fitting technique. A 3rd order Coaxial Lowpass filter and an 8th order dual-mode elliptic integral function response filter are taken as an example to validate the proposed method.

• Proceedings of the IEEK Conference
• /
• 2002.07c
• /
• pp.1626-1629
• /
• 2002

This paper describes novel methods to construct fuzzy inference rules with gradient descent. The present methods have a constructive mechanism of the rule unit that is applicable in two parameters: the central value and the width of the membership function in the antecedent part. The first approach is to create the rule unit at the nearest position from the input space, for the central value of the membership function in the antecedent part. The second is to create the rule unit which has the minimum width, for the width of the membership function in the antecedent part. Experimental results are presented in order to show that the proposed methods are effective in difference on the inference error and the number of learning iterations.

• Korean Management Science Review
• /
• v.15 no.1
• /
• pp.77-85
• /
• 1998

In this paper a face optimization algorithm is developed for solving the problem (P) of optimizing a linear function over the set of efficient solutions of a multiple objective linear program. Since the efficient set is in general a nonconvex set, problem (P) can be classified as a global optimization problem. Perhaps due to its inherent difficulty, relatively few attempts have been made to solve problem (P) in spite of the potential benefits which can be obtained by solving problem (P). The algorithm for solving problem (P) is guaranteed to find an exact optimal or almost exact optimal solution for the problem in a finite number of iterations.