• Proceedings of the KIEE Conference
• /
• 2007.07a
• /
• pp.329-330
• /
• 2007

본 논문에서는 particle swarm optimization(PSO)를 통한 비선형시스템의 퍼지집합 퍼지모델의 최적화 방법을 제안한다. 퍼지 모델링에서 전반부 동정, 즉 구조 동정 및 파라미터 동정은 비선형 시스템을 표현하는데 있어서 매우 중요하다. 퍼지모델의 전반부 동정에 있어 최적화 과정이 필요하며 유전자 알고리즘(Genetic Algorithm; GA)을 이용하여 퍼지모델을 최적화한 연구가 많이 있다. 본 연구는 파라미터 동정 시 최근 여러 가지 어려운 최적화 문제를 수행함에 있어서 성능의 우수성이 증명된 PSO를 이용하여 퍼지집합 퍼지모델의 전반부 파라미터를 동정하였다. 구조동정은 단순 유전자 알고리즘(Simple Genetic Algorithm; SGA)을 이용하여 동정하였으며 파라미터 동정시 실수 코딩유전자 알고리즘(Real Coded Genetic Algorithm; RCGA)와 PSO를 각각 파라미터 동정에 이용하여 성능을 비교하였다.

• Proceedings of the KIEE Conference
• /
• 2007.07a
• /
• pp.1815-1816
• /
• 2007

본 논문에서는 데이터의 특성을 이용한 정보 입자 기반 퍼지 뉴럴 네트워크의 연속적 최적화를 제안한다. 데이터들간의 거리를 중심으로 C-Means 클러스터링 알고리즘을 이용하여 멤버쉽 함수를 정의하고 각 중심의 후반부 중심값을 이용하여 후반부 학습에 적용한다. 구조/파라미터 동정에 있어서 실수 코딩 기반 유전자 알고리즘을 이용하여 입력변수의 수, 입력 변수의 선택, 멤버쉽함수의 수, 후반부 형태와 같은 시스템의 입력 구조와 전반부 멤버쉽함수의 정점 및 학습율과 모멘텀 계수와 같은 파라미터를 최적으로 동정한다. 또한, 구조 연산과 파라미터 연산의 연속적 동조 방법을 이용하여 퍼지 뉴럴 네트워크를 최적화한다. 제안된 퍼지 뉴럴 네트워크는 삼각형 멤버쉽 함수를 이용하며, 후반부 추론에는 간략, 선형, 변형된 2차식을 이용한다. 제안된 퍼지 뉴럴 네트워크는 표준 모델로서 널리 사용되는 수치적인 예를 통하여 평가한다.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.11 no.6
• /
• pp.876-885
• /
• 2000

A new and simple method is presented for determining the parasitic resistances of MESFET and HEMT from the measured S-parameters under normal active bias without depending on additional DC measurements or iteration or optimization process. The presented method is based on the fact that the difference between source resistance(Rs) and drain resistance(Rd) can be obtained from the measured Z-parameters under zero bias condition. It is possible to define the new internal device including intrinsic device and 3 parasitic resistances by elimination the parasitic inductances and capacitances from the measured S-parameters. Three parasitic resistances are calculated easily from the fact that the real parts of Yint,11 and Yint,12 of intrinsic Y-parameters are zero theoretically and the relations between S-,Z-, Y-matrices. The calculated parasitic resistances using the presented method and successively calculated equivalent circuit parameters give modeled S-parameters which are in good agreement with the measured S-parameters up to 400Hz.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.27 no.2B
• /
• pp.137-144
• /
• 2002

In this paper, we optimize the base station placement and transmission power using genetic approach. A new representation describing base station placement and transmit power with real number is proposed, and new genetic operators are introduced. This new representation can describe the locations, powers, and number of base stations, Considering coverage, power and economy efficiency, we also suggest a weighted objective function. Our algorithm is applied to an obvious optimization problem, and then it is verified. Moreover, our approach is tried in inhomogeneous traffic distribution. Simulation result proves that the algorithm enables to fad near optimal solution according to the weighted objective function.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.19 no.1
• /
• pp.62-68
• /
• 2009

Two new generation gap models with modified parent-centric recombination(PCX) operator are proposed. First, the self-adaptation generation gap(SGG) model is a control method that keeps a replaced probability of parents by offspring to a certain level which obtains better performance. Second, virtual cluster generation gap(VCGG) is provided to extend distances among parents using clustering, which causes it to diversify individuals. In this model, distances among parents can be controlled by size of clusters. To demonstrate the effectiveness of our two proposed approaches, experiments for three standard test problems are executed and compared to most competing current approaches, CMA-ES and Generalized Generation Gap(G3) with PCX. It is shown two proposed methods are superior to consistently other approaches in the study.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.44 no.11
• /
• pp.989-996
• /
• 2016

This paper intends to develop the rotor track and balance (RTB) algorithm using the nonlinear RTB models and a real-coded hybrid genetic algorithm. The RTB response data computed using the trim solutions with variation of the adjustment parameters have been used to build nonlinear RTB models based on the quadratic interpolation functions. Nonlinear programming problems to minimize the track deviations and the airframe vibration responses have been formulated to find optimum settings of balance weights, trim-tab deflections, and pitch-link lengths of each blade. The results are efficiently resolved using the real-coded genetic algorithm hybridized with the particle swarm optimization techniques for convergence acceleration. The nonlinear RTB models and the optimized RTB parameters have been compared with those computed using the linear models to validate the proposed techniques. The results showed that the nonlinear models lead to more accurate models and reduced RTB responses than the linear counterpart.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.2
• /
• pp.552-557
• /
• 2000

In the paper, we proposed a face recognition method that uses GA-BP(Genetic Algorithm-Back propagation Network) that optimizes initial parameters such as bias values or weights. Each pixel in the picture is used for input of the neuralnetwork. The initial weights of neural network is consist of fixed-point real values and converted to bit string on purpose of using the individuals that arte expressed in the Genetic Algorithm. For the fitness value, we defined the value that shows the lowest error of neural network, which is evaluated using newly defined adaptive re-learning operator and built the optimized and most advanced neural network. Then we made experiments on the face recognition. In comparison with learning convergence speed, the proposed algorithm shows faster convergence speed than solo executed back propagation algorithm and provides better performance, about 2.9% in proposed method than solo executed back propagation algorithm.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.04a
• /
• pp.341-344
• /
• 2007

본 논문에서는 계층적 공정 경쟁 개념을 병렬 유전자 알고리즘에 적용하여 계층적 공정 경쟁 기반 병렬유전자 알고리즘 (Hierarchical Fair Competition Genetic Algorithm: HFCGA)을 구현하였을 뿐만 아니라 실수코딩 유전자 알고리즘(Real-Coded Genetic Algorithm: RCGA)에서 좋은 성능을 갖는 산술교배(Arithmetic crossover), 수정된 단순교배(modified simple crossover) 그리고 UNDX(unimodal normal distribution crossover)등의 다양한 교배연산자들을 적용, 분석함으로써 개선된 병렬 유전자 알고리즘을 제안하였다. UNDX연산자는 다수의 부모(multiple parents)를 이용하여 부모들의 기하학적 중심(geometric center)에 근접하게 정규분포를 이루며 생성된다. 본 논문은 UNDX를 이용한 HFCGA모델을 구현하고 함수파라미터 최적화 문제에 많이 쓰이는 함수들에 적용시킴으로써 그 성능의 우수성을 증명 한다.

• Journal of Advanced Marine Engineering and Technology
• /
• v.34 no.1
• /
• pp.109-115
• /
• 2010

This paper deals with trajectory control of computer simulated mobile robot via fuzzy control. Mobile robot is controlled by Mamdani type fuzzy controller. Inputs of the fuzzy controller are angle between mobil robot and target, changed angle and output is the steering angle, which is control input. Fuzzy rules have seven rules and are selected by human experiential knowledge. Also we propose a scaling factors tuning scheme which is the another focus in designing fuzzy controller. In this paper, we adapt the RCGA which is well known in parameter optimization to adjust scaling factors. The simulation results show that the fuzzy control effectively realize trajectory stabilization of the mobile robot along a given reference target from various initial steering angles.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.2
• /
• pp.181-189
• /
• 2013

In this paper, we introduce the fuzzy neural network based on the individual input space to design the pattern recognizer. The proposed networks configure the network by individually dividing each input space. The premise part of the networks is independently composed of the fuzzy partition of individual input spaces and the consequence part of the networks is represented by polynomial functions. The learning of fuzzy neural networks is realized by adjusting connection weights of the neurons in the consequent part of the fuzzy rules and it follows a back-propagation algorithm. In addition, in order to optimize the parameters of the proposed network, we use real-coded genetic algorithms. Finally, we design the optimized pattern recognizer using the experimental data for pattern recognition.