http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
신성효,김상운 대한전자공학회 1998 電子工學會論文誌, C Vol.c35 No.2
Structural learning methods of MLP classifiers for a given application using genetic algorithms have been studied. In the methods, however, the search space for an optimal structure is increased exponentially for the physical application of high diemension-multi calss. In this paperwe propose a method of MLP classifiers using species genetic algorithm(SGA), a modified GA. In SGA, total search space is divided into several subspaces according to the number of hidden units. Each of the subdivided spaces is called "species". We eliminate low promising species from the evoluationary process in order to reduce the search space. experimental results show that the proposed method is more efficient than the conventional genetic algorithm methods in the aspect of the misclassification ratio, the learning rate, and the structure.structure.
Cubic B-Spline을 이용한 측면 얼굴의 인식 방법
신성효,이인순,김상운 명지대학교 공학기술연구소 1994 공학기술연구소 논문집 Vol.9 No.-
Profile contours can be a useful tool for facial recognition problems, and many efforts have been made in this field. These researches mainly concentrated on the feature extraction by making use of the positional information of a nose tip. Most importantly, incorrect extraction of the positional information of the nose tip degrades the overall performance of the facial recognition system. For instance, when facial images are tippd, the possibility of erroneous extration of nose tip position is high. However, endeavors to account for the tipped images. such as gathering and analyzing facial images from various viewpoints, have never been made as yet. This paper provides a method for the facial recognition by extracting multiple control points. Given a profile image, we extract the profile contour through binary processing. Since the contour is a space curve, curvature can readily be obtained at any point lying on the curve. By examining the signs at eight curvature extrimities, facial direction of the given profile image can be determined, and this information is used to derive the direction of front image. Finally, a cubic-polynomial curve which minimize the least square error at extremities is calculated and used to get control points of the corresponding cubic B-Spline. Thirteen features from the control points show the recognition rate of approximately 90 percents.
신성효,김상운 明知大學校 産業技術硏究所 1997 産業技術硏究所論文集 Vol.16 No.-
In this paper a structural learning method of optimal MLP classifiers using hybrid methods is proposed. The structural learning methods are catagorized 3 groups. The first group is "entropy factor" or " sequential network creation" that determine the number of hidden neurons. Second is "sensitivity based pruning" that remove the unnecessary input neuron, and the last is "entropy minimum learning" that remove the unnecessary links. To evaluate the hybrid methods we experiment with two kinds of data set : 4 dimension's IRIS, 63 dimension's E13B. Experimental results show that the methods could generate a task-related MLP structure which does not contain unimportant nodes and links.