• 한국정보통신학회논문지
• /
• 제12권9호
• /
• pp.1682-1688
• /
• 2008

CP(Counterpropagation) 알고리즘은 Kohonen의 경쟁 네트워크와 Grossberg의 아웃스타(Outstar) 구조의 결합으로 이루어진 것으로 패턴 매칭, 패턴 분류, 통계적인 분석 및 데이터 압축 등 활용분야가 다양하고, 다른 신경망 모델에 비해 학습이 매우 빠르다는 장점이 있다. 그러나 CP 알고리즘은 충분한 경쟁층의 수가 설정되지 않아 경쟁층에서 학습이 불안정하고, 다양한 패턴으로 구성된 경우에는 패턴들을 정확히 분류할 수 없는 경우가 발생한다. 그리고 CP 알고리즘은 출력층에서 연결 강도를 조정할 때, 학습률에 따라 학습 및 인식 성능이 좌우된다. 본 논문에서는 효과적인 패턴인식을 위해 다수 경쟁층을 설정하고, 입력 벡터와 승자 뉴런의 대표 벡터간의 차이와 승자 뉴런의 빈도수를 학습률 조정에 반영하고 학습률을 동적으로 조정하여 경쟁층에서 안정적으로 학습되도록 하고, 출력층의 연결강도를 조정할 때 모멘텀(Momentum) 방법을 적용한다. 제안된 CP 학습 성능을 확인하기 위해서 실제 여권에서 추출된 개별 코드를 대상으로 실험한 결과, 개선된 CP 알고리즘이 기존의 CP 알고리즘보다 학습 성능, 분류의 정확성 및 인식 성능이 개선된 것을 확인하였다.

• 대한원격탐사학회지
• /
• 제38권6_2호
• /
• pp.1589-1605
• /
• 2022

위상 언래핑은 위성레이더 간섭기법의 필수적인 자료처리 절차다. 이에 따라 비 딥러닝 기반 언래핑 기법이 다수 개발되었으며 최근에는 딥러닝 기반 언래핑 기법이 제안되고 있다. 본 논문에서는 딥러닝 기반 위성레이더 언래핑 기법을 1) 언래핑된 위상의 예측 방법, 2) 위상 언래핑을 위한 딥러닝 모델의 구조 그리고 3) 학습데이터 제작 방법의 측면에서 최근 연구 동향을 소개하였다. 언래핑된 위상을 예측하는 방법은 모호 정수 분류방법, 위상 단절 구간 탐지 방법, 위상 예측 방법, 딥러닝과 전통적인 언래핑 기법의 연계 방법에 따라 다시 세분화하여 연구 동향을 나타냈다. 일반적으로 활용되는 딥러닝 모델 구조의 특징과 전체 위상 정보를 파악하기 위한 모델 최적화 방법에 대한 연구 사례를 소개하였다. 또한 학습데이터 제작 방법은 주로 위상 변이 제작과 노이즈 시뮬레이션 방법으로 구분하여 연구 동향을 정리하였으며 추후 발전 방향을 제시하였다. 본 논문이 추후 국내의 딥러닝 기반 위상 언래핑 연구의 발전 방향을 모색하는 데에 필요한 기반 자료로 활용되기를 기대한다.

• 농업과학연구
• /
• 제49권3호
• /
• pp.619-627
• /
• 2022

In Korea, apples and pears are among the most important agricultural products to farmers who seek to earn money as income. Generally, farmers make decisions at various stages to maximize their income but they do not always know exactly which option will be the best one. Many previous studies were conducted to solve this problem by predicting farmers' income structure, but researchers are still exploring better approaches. Currently, machine learning technology is gaining attention as one of the new approaches for farmers' income prediction. The machine learning technique is a methodology using an algorithm that can learn independently through data. As the level of computer science develops, the performance of machine learning techniques is also improving. The purpose of this study is to predict the income structure of apples and pears using the automatic machine learning solution H2O.AI and to present some implications for apple and pear farmers. The automatic machine learning solution H2O.AI can save time and effort compared to the conventional machine learning techniques such as scikit-learn, because it works automatically to find the best solution. As a result of this research, the following findings are obtained. First, apple farmers should increase their gross income to maximize their income, instead of reducing the cost of growing apples. In particular, apple farmers mainly have to increase production in order to obtain more gross income. As a second-best option, apple farmers should decrease labor and other costs. Second, pear farmers also should increase their gross income to maximize their income but they have to increase the price of pears rather than increasing the production of pears. As a second-best option, pear farmers can decrease labor and other costs.

• 공학교육연구
• /
• 제18권2호
• /
• pp.77-88
• /
• 2015

Recently, there has been an increasing interest in 'Flipped Learning,' an IT-based learner-centered teaching-learning method corresponding to meet the paradigm of the future education. For smooth Flipped Learning, there are three steps in total: a pre-class should precede; then, in the structure of classes in the classroom, in-class learning among peer learners should be done; and lastly, the operation of a post-class should be done. For successful Flipped Learning, class elements in each step should be designed with a time difference, interconnected so as to achieve a single educational objective. However, it was found that there was a limitation in that the teaching-learning model of the preceding Flipped Learning consisted of the order of analysis, design, development, implementation and evaluation as general procedures, so it would not sufficiently consider the situations of Flipped Learning only. On this background, this thesis proposes a differentiated Flipped Learning model for mastery learning in a subject of an institute of technology as a model of systematic instructional design and presents a case of a class applied to an actual subject of computer engineering.

• International Journal of Control, Automation, and Systems
• /
• 제6권5호
• /
• pp.755-766
• /
• 2008

This study presents a recurrent compensatory fuzzy neural network (RCFNN) for dynamic system identification. The proposed RCFNN uses a compensatory fuzzy reasoning method, and has feedback connections added to the rule layer of the RCFNN. The compensatory fuzzy reasoning method can make the fuzzy logic system more effective, and the additional feedback connections can solve temporal problems as well. Moreover, an online learning algorithm is demonstrated to automatically construct the RCFNN. The RCFNN initially contains no rules. The rules are created and adapted as online learning proceeds via simultaneous structure and parameter learning. Structure learning is based on the measure of degree and parameter learning is based on the gradient descent algorithm. The simulation results from identifying dynamic systems demonstrate that the convergence speed of the proposed method exceeds that of conventional methods. Moreover, the number of adjustable parameters of the proposed method is less than the other recurrent methods.

• 한국멀티미디어학회논문지
• /
• 제23권9호
• /
• pp.1129-1138
• /
• 2020

The performance of convolutional deep learning networks is generally determined according to parameters of target dataset, structure of network, convolution kernel, activation function, and optimization algorithm. In this paper, a genetic algorithm is used to select the appropriate deep learning model and parameters for Alzheimer's classification and to compare the learning results with preliminary experiment. We compare and analyze the Alzheimer's disease classification performance of VGG-16, GoogLeNet, and ResNet to select an effective network for detecting AD and MCI. The simulation results show that the network structure is ResNet, the activation function is ReLU, the optimization algorithm is Adam, and the convolution kernel has a 3-dilated convolution filter for the accuracy of dementia medical images.

• 한국어교육
• /
• 제28권4호
• /
• pp.297-335
• /
• 2017

Recently, Chinese universities have been putting their priority on cultivating industry-academia linked talents, catching up with social change and industrial demand. Accordingly, vocational purpose education is being emphasized even in Korean language education. When facing active trade between Korea and China, the importance of business Korean language education will be magnified, and therefore, the demand for business Korean textbooks will grow accordingly. To strengthen the basis for the development of future business Korean textbooks, this study conducted a general analysis of a business Korean textbook for Chinese learners. Specifically, the textbook was examined by largely dividing it into external and internal structures. After dividing the internal structure into "purpose of compilation", "composition of the textbook", "learning contents", and "learning activity", the composition of the textbook was once again divided into "overall composition" and "unit composition", and the learning contents was further divided into "subject", "language content", and "supplementary knowledge." Furthermore, an analysis was conducted. The status and directions for future development of business Korean textbooks for Chinese learners are delineated and suggestions for improvement are provided. This study has its significant in that a general analysis was conducted on a business Korean textbook for Chinese learners, and is expected to be used as basic research material for the future development of business Korean textbooks.

• 한국공간구조학회논문집
• /
• 제21권2호
• /
• pp.41-48
• /
• 2021

A smart tuned mass damper (TMD) is widely studied for seismic response reduction of various structures. Control algorithm is the most important factor for control performance of a smart TMD. This study used a Deep Deterministic Policy Gradient (DDPG) among reinforcement learning techniques to develop a control algorithm for a smart TMD. A magnetorheological (MR) damper was used to make the smart TMD. A single mass model with the smart TMD was employed to make a reinforcement learning environment. Time history analysis simulations of the example structure subject to artificial seismic load were performed in the reinforcement learning process. Critic of policy network and actor of value network for DDPG agent were constructed. The action of DDPG agent was selected as the command voltage sent to the MR damper. Reward for the DDPG action was calculated by using displacement and velocity responses of the main mass. Groundhook control algorithm was used as a comparative control algorithm. After 10,000 episode training of the DDPG agent model with proper hyper-parameters, the semi-active control algorithm for control of seismic responses of the example structure with the smart TMD was developed. The simulation results presented that the developed DDPG model can provide effective control algorithms for smart TMD for reduction of seismic responses.

• 한국건축시공학회:학술대회논문집
• /
• 한국건축시공학회 2021년도 가을 학술논문 발표대회
• /
• pp.223-224
• /
• 2021

The process before the model learning stage in AI R&D can be subdivided into data collection/cleansing-data purification-data labeling. After that, according to the purpose of development, it goes through a stage of verifying the model by performing learning by using the algorithm of the artificial intelligence model. Several studies describe an important part of AI research as the learning stage, and try to increase the accuracy by changing the structure and layer of the AI model. However, if the refinement and labeling process of the learning data is tailored only to the model format and is not made for the purpose of development, the desired AI model cannot be obtained. The latest research reveals that most AI research failures are the failure of the learning data rather than the structure of the AI model. analyzed.

• 실천공학교육논문지
• /
• 제15권1호
• /
• pp.223-232
• /
• 2023

일학습병행은 산업현장과 학교 교육의 미스매치를 해소하고 능력중심 사회를 구현하기 위한 핵심적인 정책으로 추진되어 2022년 12월 기준으로 16,664개 기업이 훈련에 참여하였다. 학습기업은 현장훈련을 실시하는 교육훈련 공급기관으로써 매우 중요한 역할을 수행하고 있다. 본 연구에서는 일학습병행에 참여하고 있는 학습기업의 평가를 위해 해당기업을 담당하고 있는 전문가들의 인지구조 분석을 통해 기업현장 교육훈련의 질을 결정하는 중요 요인을 도출하고 학습기업의 평가지표를 제시하여 일학습병행의 질적 내실화를 도모하는 기초 자료로 활용될 것이다.