Go to the main menu
Skip to content
Go to bottom
REFERENCE LINKING PLATFORM OF KOREA S&T JOURNALS
> Journal Vol & Issue
Journal of Korean Institute of Intelligent Systems
Journal Basic Information
Journal DOI :
Korean Institute of Intelligent Systems
Editor in Chief :
Volume & Issues
Volume 9, Issue 6 - Dec 1999
Volume 9, Issue 5 - Oct 1999
Volume 9, Issue 4 - Aug 1999
Volume 9, Issue 3 - 00 1999
Volume 9, Issue 2 - 00 1999
Volume 9, Issue 1 - 00 1999
Selecting the target year
Color Pattern Retrievals Using Linguistic Image Words
Eom, Jin Seop ; Lee, Jun Hwan ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 1~1
In this paper, a color pattern retrieval system is proposed, which recommends color patterns of thedesired feeling. The desired feeling is a 9-dimensional vector on the 9 linguistic image scales. Each component of the vector represents the degree of the corresponding image scale consisted of opposite concepts. In order to construct the system, the physical features of stored color patterns are extracted, and transformed to the 9-dimensional emotional features. In the retrieval process, the query vector Is compared with the emotional features of stored color patterns, and the system recommend the color pat-terns that have similar emotional features. To make an efficient indexing system, the hierarchical clustering method with Fuzzy c-means algorithm is used. Even though several problems are still remained, the experimental results shows the system can be used to select the wall papers, textile designs, and pictures in gallery.
Adaptive Genetic Algorithms Using Reinforcement Learning
Lee, Sang Hwan ; Jeon, Hyo Byeong ; Sim, Gwi Bo ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 10~10
Genetic Algorithms(GAs) are stochastic computational models inspired by natural phenomena. Twomajor procedures of GAs, therefore, are operation and selection. By means of operation such as crossover, mutation, inversion, etc., the individuals can produce new offsprings, and through the selection mechanism the fitter ones can produce more offspring than the less fit ones. GAs, however, have some risks to 1111 into local minima because their control parameters such as population size, crossover rate, and mutation probability, can not be obtained deterministically but heuristically. We, therefore, propose an adaptive genetic algorithm(AGA), which generate mutation probabilities of each locus by interacting with the environment according to reinforcement learning. We verify the effectiveness of the proposed method via several simulation results compared with simple GAs.
Revising the Traditional Backpropagation with the Method of Conjugate Gradients and Optimizing Learning Parameters
Choe, Sang Ung ; Lee, Jin Chun ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 17~17
In this paper, we propose a new paradigm(CGBP-Conjugate Gradient method applied to Back-Propagation) to be capable of overcoming limitations of the traditional backpropagation. The CGBP is based on the method of conjugate gradients and calculates teaming parameters through the line search which may be characterized by order statistics and golden section. Experimental results show that the CGBP is definitely superior to both the stochastic and the deterministic traditional backpropagation in terms of accuracy and rate of convergence and may surmount the problem of local minima. Furthermore, they confirm us that the stagnant phenomenon of teaming in the traditional backpropagation results from the limitations of its algorithm in itself and that unessential approaches would never cured it of this phenomenon.
Program Complexity Measurement Based Rough and Fuzzy Sets
Choe, Wan Gyu ; Na, Yeong Nam ; Kim, Yeong Sik ; Lee, Seong Ju ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 41~41
As a result of increase of size and complexity of software system, many researches about measurement of the program complexity have been in progress so as to identify and classify characteristics of software. But the complexity metrics using only a factor or the hybrid metrics using many factors without suggesting significance of each factor can't exactly measure and assess the program complexity. Therefore, in this paper, we propose the new hybrid metric which can measure the complexity of function-oriented programs by using the objective metrics proven by many researches and experiments in industry field and a significance of each metric looked upon as a weight.
A Method of Fuzzy Compensation for Stabilization of a Gait of a Quadruped Walking Robot
Lee, Yeon Jeong ; Gang, Dong O ; Lee, Seung Ha ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 48~48
In this paper, a gait design problem for a quadruped walking robot is addressed. In conventional researches, a gait is designed by a geometrical method with an assumption that the weight of legs can be ignored. However, the robot would be unstable when it walks with the designed gait, since the weight of legs is too heavy to be negligible. To resolve this problem of discrepancy, we propose a fuzzy compensator and a quadruped model for calculation of position of an effective center of gravity inconsideration of the weight of legs. It is shown by computer simulation results that the stability of the quadruped is guaranteed by the proposed method of fuzzy compensation.
Separation Axioms in Smooth Fuzzy Topological Spaces
Kim, Yong Chan ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 57~57
A Neuro-Fuzzy Approach to Integration and Control of Industrial Process : Part Ⅱ
Kim, Seong Sin ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 63~63
Modeling for Fuzzy Neural Network Using Genetic Algorithm and Its Application to Nonlinear System Modeling
Jang, Uk ; Ju, Yeong Hun ; Park, Jin Bae ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 71~71
This paper proposed the hybrid algorithm for the optimization of the structure and parameters of thefuzzy neural networks by genetic algorithms (GA) to improve the behaviour and design of fuzzy neural networks. Fuzzy neural networks have a distinguishing feature in that they can possess the advantage of both neural networks and fuzzy systems. In this way, we can bring the low-level learning and computational power of neural networks into fuzzy systems and also high-level, human like If-THEN rule thinking and reasoning of fuzzy systems into neural networks. As a result, there are many research works concerning the optimization of the structure and parameters of fuzzy neural networks. In this pap-er, we propose the hybrid algorithm that can optimize both the structure and parameters of fuzzy neural networks. The proposed method can systematically optimize fuzzy neural networks by combining traditional works and genetic algorithms. Numerical examples are provided to show the advantages of the proposed method.
Eugenic Genetic Algorithm
Jeong, Seong Hun ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 81~81
A Study on Chaotic Synchronization and Secure Communication of Chua's Circuit with Transmission Line
Bae, Yeong Cheol ; Kim, Lee Gon ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 89~89
In this paper, a transmitter and a receiver using two identical Chua's circuits are proposed and RLCGsynchronization and a wire secure communicationsa are investigated. As several problems have been found in both the drive-response synchronization and the coupled synchronization in the previous researches, a new drive-coupled synchronization theory is proposed that can be applicable to wire communication. Since the synchronization of the wire transmission system is impossible by coupled synchronization, theory having both the drive-response and the coupled synchronization is proposed. As a result the chaotic synchronization has delay characteristics in the RLCG transmission system caused by the line parameters L and C. A secure communication method in which the desired information signal is synthesized with the chaos signal created by the Chua's circuit is proposed and information signal is de-modulated also using the Chua's circuit. The proposed method is synthesizing the desired information with the chaos circuit by adding the information signal to the chaos signal in the wire transmission system. After transmitting the synthesized signal through the wire transmission system, it is confirmed the feasibility of the secure communication from result of demodulated signals and recovered wire tapped signals.
A Study Power control for Inverter Spot Welder Using Evolution Strategy
Kim, Jae Mun ; Won, Chung Yeon ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 97~97
In this paper, we attempt to control the power system using Pl controller to improve the quality ofthe welding products. Constant current control strategy used in industrial field brings about the depreciation of welding products because it makes its workpiece spattered irrespective of dynamic resistance characteristic during welding process. We also propose a Pl controller for spot welder system using ES(Evolution Strategy) with varying search space.25 with varying search space which depends on fitness values at each generation is used to tune Pl control parameters. Simulation results show that the pro-posed algorithm has high performances with effective search ability.
Moving Obstacles Collision Avoidance of a Mobile Robot using GA and Neural Network
Park, Yun Myeong ; Jang, Jong Seung ; Han, Chang Hun ; Im, Yeong Do ; Choe, Bu Gwi ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 106~106
This paper proposes a new construction method of neural networks. The construction method consists oftwo fundmental ideas, which are a parallel selection-style evaluation and rules evolution. A new collision a-voidance algorithm using genetic and neural network is proposed for avoid moving obstacles such as mobile robots. The input parameters of this algorithm is position of moving obstacles and target. Output is a regenerated direction of mobile robot. This algorithm is very simple and so, it is available to application of real time process. The pattern of collision avoidance is learned through test execution
Handwritten Numeral Character Recognition Using the Tolerant Rough Set
Kim, Dae Jin ; Kim, Cheol Hyeon ;
Journal of Korean Institute of Intelligent Systems, volume 9, issue 1, 1999, Pages 113~113
This paper proposes a new data classification method based on the tolerant rough set that extends theexisting equivalent rough set. Similarity measure between two data is described with a distance function of each attribute and two data are defined to be tolerant when they have the similarity measure that exceeds a threshold value. In this case, the optimal threshold values for defining the similar data are obtained by evolving the threshold values by genetic algorithm whose fitness function is a balance function that is not only to maximize the good connections but also to minimize bad connections. Further, these tolerant data set are classified into two approximate sets . lower and upper approximation de-pending on the coincidence of their classes. A hierarchical classification method is utilized such that all data are classified by using the lower approximation in the first stage and then the non-classified data by the lower approximation are classified again by using the rough membership functions obtained from the upper approximation. The validity of our proposed classification method is verified by ap-plying the proposed classification method to the handwritten numeral character and by comparing the classification error and teaming time with the feed forward neural network's backpropagation algorithm.