http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
인공신경망 가중치의 효과적인 최적화를 위한 하이브리드 메타휴리스틱 알고리즘
고광종,허재석 대한산업공학회 2022 대한산업공학회지 Vol.48 No.2
Due to a number of local minimums, the gradient descent methods have difficulties in searching for the optimal weights of artificial neural networks (ANNs). To resolve this problem, recently, a number of scholars have conducted studies that utilize metaheuristics to optimize the weights of ANNs. Particularly, the swarm intelligence algorithm which is one of the metaheuristics has shown great potential in previous studies. In this study, we propose the hybrid metaheuristic algorithm that effectively optimizes the weight of ANNs by combining particle swarm optimization (PSO) and grey wolf optimizer (GWO), both of which are swarm intelligence algorithms. In the search process, the proposed algorithm resolves the convergence instability in the validation dataset using the swarm memory that was inspired by the habit of remembering the environment in which the swarm was suitable for survival. Numerical experiments demonstrate the proposed algorithm outperforms the existing swarm intelligence algorithms, stochastic gradient descent, and Adam Optimizer in terms of classification accuracy.
소셜 감성 데이터를 이용한 딥러닝 기반의 암호화폐 가격 예측
고광종(Gwangjong Ko),김다솔(Dasol Kim),허재석(Jaeseok Huh) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
소셜 감성 데이터를 활용하여 암호화폐의 가격을 예측하는 기존의 연구들은 데이터 수집의 출처가 제한적이어서 다양한 사회적 분위기를 반영하기 어렵다는 한계가 존재했으며, 사전 기반 언어 모델을 사용했기 때문에 최근의 유행어를 분석할 수 없었다. 본 연구에서는 상기 한계를 보완하기 위해 다양한 출처의 소셜 감성 데이터를 딥러닝 기반 언어 모델 BERT (Bidirectional Encoder Representations from Transformers)로 분석한다. 분석이 완료된 후, 분석 결과를 시계열 종속 변수로 변환하는 본 연구의 제안 방법을 통해 비트코인의 한 시간 단위 종가를 예측한다. 실험을 통해 제안 방법은 가격 예측 오차 감소에 효과가 있는 것으로 검증되었다. Existing studies that predict the cryptocurrency price using social sentiment data have difficulty in considering various social atmospheres due to the limited sources of data collection. In addition, it was not possible to analyze buzzwords because they use a dictionary-based language model. In this study, to resolve the above problems, we analyze social sentiment data from various sources using BERT (Bidirectional Encoder Representations from Transformers), a deep learning-based language model. Then, the proposed method converts the result of the sentiment analysis into time-series independent variables to predict the hourly closing price of Bitcoin. Numerical experiments demonstrate the proposed method is effective in the reduction of price prediction error.
초등학생의 태권도 수련활동 참가에 따른 참여동기와 수련만족의 관계
성영호+고광종 용인대학교 2006 용인대학교 논문집 Vol.24 No.-
The purpose of this research was establish a hypothetical cause-effect model based on the precedent works related to motive for participation and training satisfaction according to Taekwondo drill of elementary school children and to investigate relations there of. To test this, elementary school children living in the vicinity of Seoul as of 2005 were selected as population and 280 children were asked to complete questionnaire and the data were analyzed by Stratified Random Cluster Sampling. Cronbach's Alpha, reliability of questionnaire used in this survey was .556~. 725, so then it showed higher internal consistency. Analytical techniques used in this research work were Frequency Analysis, Reliability Analysis, T-Test, oneway ANOVA, and Multi Regression Analysis. Results from these analyses revealed the followings: First, results from analysis of motives for participation according to gender revealed that there were statistically significant differences in intellectual motivation, social motivation, cognitive motivation, and of training satisfaction that there were statistically significant differences in physical satisfaction and mental satisfaction. Regarding motive for participation according to year, there were no statistically significant differences. And results from analysis of training satisfaction indicated that there were statistically significant differences in satisfaction of acquiring skill, physical satisfaction, and mental satisfaction, and the motives for participation according to monthly income revealed statistically significant difference in intellectual motivation, and also training satisfaction revealed statistically significant difference in physical satisfaction. 성 영 호․고 광 종 초등학생의 태권도 수련활동 참가에 따른 참여동기와 수련만족의 관계 Second, results from the analysis of motive for participation according to the level of training showed that there were statistically significant differences in intellectual motivation, social motivation and cognitive motivation, and training satisfaction showed statistically significant differences in satisfaction of acquiring skill and mental satisfaction. And motive for participation according to duration of training revealed statistically significant differences in intellectual motivation, social motivation and cognitive motivation, and there were no statistically significant differences in job satisfaction. Third, results from analysis of level of participation and duration of participation exhibited that these two factors had no influences on motive for participation and training satisfaction. Fourth, results from analysis showed that intellectual motivation and cognitive motivation had influences on satisfaction of self-defense, and that intellectual motivation and cognitive motivation affected satisfaction of acquiring skill, and that cognitive motivation had impact physical satisfaction. And, that all variables including intellectual motivation, social motivation and cognitive motivation had influences on mental satisfaction.