• Journal of Korea Multimedia Society
• /
• v.25 no.10
• /
• pp.1375-1385
• /
• 2022

In this paper, we propose a semi-supervised domain adaptation solution to deal with practical face recognition (FR) scenarios where a single face image for each target identity (to be recognized) is only available in the training phase. Main goal of the proposed method is to reduce the discrepancy between the target and the source domain face images, which ultimately improves FR performances. The proposed method is based on the Domain Adatation network (DAN) using an MMD loss function to reduce the discrepancy between domains. In order to train more effectively, we develop a novel loss function learning strategy in which MMD loss and cross-entropy loss functions are adopted by using different weights according to the progress of each epoch during the learning. The proposed weight adoptation focuses on the training of the source domain in the initial learning phase to learn facial feature information such as eyes, nose, and mouth. After the initial learning is completed, the resulting feature information is used to training a deep network using the target domain images. To evaluate the effectiveness of the proposed method, FR performances were evaluated with pretrained model trained only with CASIA-webface (source images) and fine-tuned model trained only with FERET's gallery (target images) under the same FR scenarios. The experimental results showed that the proposed semi-supervised domain adaptation can be improved by 24.78% compared to the pre-trained model and 28.42% compared to the fine-tuned model. In addition, the proposed method outperformed other state-of-the-arts domain adaptation approaches by 9.41%.

• Smart Structures and Systems
• /
• v.30 no.6
• /
• pp.687-701
• /
• 2022

Inevitable response loss under complex operational conditions significantly affects the integrity and quality of measured data, leading the structural health monitoring (SHM) ineffective. To remedy the impact of data loss, a common way is to transfer the recorded response of available measure point to where the data loss occurred by establishing the response mapping from measured data. However, the current research has yet addressed the structural condition changes afterward and response mapping learning from a small sample. So, this paper proposes a novel data driven structural response reconstruction method based on a sophisticated designed generating adversarial network (UAGAN). Advanced deep learning techniques including U-shaped dense blocks, self-attention and a customized loss function are specialized and embedded in UAGAN to improve the universal and representative features extraction and generalized responses mapping establishment. In numerical validation, UAGAN efficiently and accurately captures the distinguished features of structural response from only 40 training samples of the intact structure. Besides, the established response mapping is universal, which effectively reconstructs responses of the structure suffered up to 10% random stiffness reduction or structural damage. In the experimental validation, UAGAN is trained with ambient response and applied to reconstruct response measured under earthquake. The reconstruction losses of response in the time and frequency domains reached 16% and 17%, that is better than the previous research, demonstrating the leading performance of the sophisticated designed network. In addition, the identified modal parameters from reconstructed and the corresponding true responses are highly consistent indicates that the proposed UAGAN is very potential to be applied to practical civil engineering.

• Journal of the Semiconductor & Display Technology
• /
• v.22 no.3
• /
• pp.112-118
• /
• 2023

As technology advances, the need for enhanced preparedness against cyber-attacks becomes an increasingly critical problem. Therefore, it is imperative to consider various circumstances and to prepare for cyber-attack strategic technology. This paper proposes a method to solve network security problems by applying reinforcement learning to cyber-security. In general, traditional static cyber-security methods have difficulty effectively responding to modern dynamic attack patterns. To address this, we implement cyber-attack scenarios such as 'Tiny Alpha' and 'Small Alpha' and evaluate the performance of various reinforcement learning methods using Network Attack Simulator, which is a cyber-attack simulation environment based on the gymnasium (formerly Open AI gym) interface. In addition, we experimented with different RL algorithms such as value-based methods (Q-Learning, Deep-Q-Network, and Double Deep-Q-Network) and policy-based methods (Actor-Critic). As a result, we observed that value-based methods with discrete action spaces consistently outperformed policy-based methods with continuous action spaces, demonstrating a performance difference ranging from a minimum of 20.9% to a maximum of 53.2%. This result shows that the scheme not only suggests opportunities for enhancing cybersecurity strategies, but also indicates potential applications in cyber-security education and system validation across a large number of domains such as military, government, and corporate sectors.

• Journal of Distribution Science
• /
• v.16 no.7
• /
• pp.67-75
• /
• 2018

Purpose - In the 21st century, as information technology advances alongside the emergence of the 4th generation, industrial age, industrial environment has become individualized and customized. It is important to hire good quality employees for good service in the industry. The e-learning market is growing every year. Although e-learning companies are finding better quality employees in e-learning, it is not easy to find it. Companies also spend a lot of time and cost to find employee. On the employees side, they want to get a job freely when they want, but they cannot find their job easily. Furthermore, the labor market environment is changing fast. In the 4th generation, industrial age, employers require to find manpower whenever they need and want at little cost. So of their own accord, we have considered the necessity of management of human resources for employees and employers in e-learning. The purpose of this study is to propose a human cloud platform framework for enabling an efficient management of human resources in e-learning industry. Research design, data, and methodology - To pinpoint the items of a human cloud platform framework, the study was initiated according to the following process. First, items of competency relating to e-learning instructional designer was analyzed. Second, based on the items of information from this analysis, selection and validity verification took place with 5 e-learning specialists group. Third, the opinion of experts who were in charge of hiring in e-learning companies were collated with the questionnaire. Lastly, the human cloud platform framework was proposed based on opinion results. Results - The framework was comprised of 7 domains and 27 items in order to develop the human cloud platform for e-learning instructional designer. The analysis results showed that the most highly considered item were 'skill (4.60)' that employee already have the capability. Following this (in order) were 'project type (4.56)', 'work competency (4.56)', and 'strength area of instructional design (4.52)'. Conclusions - The 27 items in the human cloud platform framework were suggested in this study. Following this, we can consider to develop the human cloud platform for finding a job and hiring e-learning instructional designer easily. For successful platform operation, we need to consider reliability between employer and employee. In addition, we need quality assurance system based on operation has public confidence.

• Journal of Internet Computing and Services
• /
• v.20 no.4
• /
• pp.21-27
• /
• 2019

Selection of feature pattern gathered from the observation of the RNA sequencing data (RNA-seq) are not all equally informative for identification of differential expressions: some of them may be noisy, correlated or irrelevant because of redundancy in Big-Data sets. Variable selection of feature pattern aims at differential expressed gene set that is significantly relevant for a special task. This issues are complex and important in many domains, for example. In terms of a computational research field of machine learning, selection of feature pattern has been studied such as Random Forest, K-Nearest and Support Vector Machine (SVM). One of most the well-known machine learning algorithms is SVM, which is classical as well as original. The one of a member of SVM-criterion is Support Vector Machine-Recursive Feature Elimination (SVM-RFE), which have been utilized in our research work. We propose a novel algorithm of the SVM-RFE with Q-learning in reinforcement learning for better variable selection of feature pattern. By comparing our proposed algorithm with the well-known SVM-RFE combining Welch' T in published data, our result can show that the criterion from weight vector of SVM-RFE enhanced by Q-learning has been improved by an off-policy by a more exploratory scheme of Q-learning.

• The Journal of the Korea Contents Association
• /
• v.21 no.1
• /
• pp.363-372
• /
• 2021

Aim of this study was to confirm the effect of teaching and learning methods on outcomes of learner according to learning style. For this, 22 of dental hygiene students(case group) was treated teaching & learning methods according to learning style while 24 of students(control group) was non treated. Pre-survey were performed before performance of program. Formative Evaluation(FE) was conducted in 2, 3 and 4 week of program respectively and summative evaluation(SE), survey of subject interest(SI) and learning motivation(LM) were conducted in 5 week. The result of study, FE, SI and LM after treatment were increased than before treatment in case group(p<0.05). SI and LM of case group were higher than control group(p<0.05). FE after treatment was increased than before treatment in he assimilator(p<0.05). SI and LM of case groups were higher than control group in assimilator and diverger(p<0.05). The result of correlation analysis, SI was related with SE, FE, LM(p<0.01, p<0.05). Thus, it is necessary to development, application and study of teaching & learning consider to learning style.

• Journal of the Korean Society for information Management
• /
• v.34 no.4
• /
• pp.7-32
• /
• 2017

Recently, deep learning has been rapidly spreading as an innovative machine learning technique in various domains. This study explored the research trends of deep learning via modified ego centered topic citation analysis. To do that, a few seed documents were selected from among the retrieved documents with the keyword 'deep learning' from Web of Science, and the related documents were obtained through citation relations. Those papers citing seed documents were set as ego documents reflecting current research in the field of deep learning. Preliminary studies cited frequently in the ego documents were set as the citation identity documents that represents the specific themes in the field of deep learning. For ego documents which are the result of current research activities, some quantitative analysis methods including co-authorship network analysis were performed to identify major countries and research institutes. For the citation identity documents, co-citation analysis was conducted, and key literatures and key research themes were identified by investigating the citation image keywords, which are major keywords those citing the citation identity document clusters. Finally, we proposed and measured the citation growth index which reflects the growth trend of the citation influence on a specific topic, and showed the changes in the leading research themes in the field of deep learning.

• Journal of The Korean Association For Science Education
• /
• v.35 no.3
• /
• pp.375-382
• /
• 2015

Textbooks corresponding to curriculum goals are necessary because they are specific products of curriculum and are the most important materials for teaching, learning, and evaluation. In particular, learning objectives written in textbooks should be clearly described because they play a role in promoting learning by showing learning goals to learners clearly. This study analyzed the characteristics of verbs used as predicate of learning objectives written in high school physics I and II textbooks of Korea and compared them with physics textbooks of Singapore and the United States. Results show that Korean textbooks have less kinds of verbs compared to those of Singapore and the United States, and the verbs with abstract and comprehensive meaning such as 'understand' and 'know' were mainly used. In American textbooks, it was noticeable that no verbs have been used by more than 10%. When classifying the learning objectives in the two Korean textbooks, cognitive domain accounted for 98 to 99%, and inquiry domain accounted for only 1% to 2%. With regard to physics textbooks of the United States, inquiry domain accounted for a large proportion of domains in learning objectives compared with physics textbooks of Korea and Singapore. Physics textbooks of Singapore were similar to those of Korea in that learning objectives were biased toward cognitive domain, but differed from those of Korea in that learning objectives were specifically described using action verbs.

• Hwankyungkyoyuk
• /
• v.7 no.1
• /
• pp.1-17
• /
• 1994

This study was performed to determine appropriate teaching-learning and evaluation methods for Environmental Studies. To promote the relevance of our study to the needs of the schools and concerned educational communities of environmental education, we reviewed related literature, conducted questionnaire surveys, interviewed related teachers and administrator, held meetings with experts, and field-tested our findings. For selecting and developing teaching-learning methods of Environmental Studies, findings of educational research in general are considered. moreover, principles of environmental education, general aim of environmental education, orientations of environmental education, and developmental stages of middle school students in educational psychology were attended. In addition, relevance to the purpose of the Environmental Studies curriculum, appropriateness for value inquiry as well as knowledge inquiry, small group centered class organization, social interaction centered teaching-learning process, regional environmental situation, significance of personal environment, evaluation methods of Environmental Studies, multi- and inter-disciplinary contents of the Environmental Studies textbook, suitability to the evaluation methods of Environmental Studies, and emphasis on the social interaction in teaching-learning process were regarded. It was learned the Environmental Studies can be taught most effectively in via of holding discussion sessions, conducting actual investigation, doing experiment-practice, doing games and plate, role-playing and carrying out simulation activities, and doing inquiry. These teaching-learning methods were field-tested and proved appropriate methods for the subject. For selecting and developing evaluation method of Environmental Studies, such principles and characteristics of Environmental Studies as objective domains stated in the Environmental Studies curriculum, diversity of teaching-learning organization, were appreciated. We categorized nine evaluation methods: the teacher may conduct questionnaire surveys, testings, interviews, non-participatory observations; they may evaluate student's experiment-practice performances, reports preparation ability, ability to establish a research project, the teacher may ask the students to conduct a self-evaluation, or reciprocal evaluation. To maximize the effect of these methods, we further developed an application system. It considered three variables, that is, evaluates, evaluation objectives domains, and evaluation agent, and showed how to choose the most appropriate methods and, when necessary, how to combine uses of different methods depending on these variables. A sample evaluation instrument made on the basis of this application system was developed and tested in the classes. The system proved effective. Pilot applications of the teaching-learning methods and evaluation method were made simultaneously; and the results and their implications are as follows. Discussion program was applied in a lesson dealing with the problems of waste disposal, in which students showed active participation and creative thinking. The evaluation method used in this lesson was a multiple-choice written test for knowledge and skills. It was shown that this evaluation method and device are effective in helping students' revision of the lesson and in stimulating their creative interpretations and responces. Pupils showed great interests in the actual investigation program, and this programme was proved to be effective in enhancing students' participation. However, it was also turned out that there must be pre-arranged plans for the objects, contents and procedures of survey if this program is to effective. In this lesson, non-participatory observation methods were used with a focus on the attitudes of students. A scaled reported in general description rather than in grade. Experiment-practice programme was adopted in a lesson for purifying contaminated water and in this lesson, instruction objectives were properly established, the teaching-learning process was clearly specified and students were highly motivated. On the other hand, however, it was difficult to control the class when some groups of students require more times to complete their experiment, and sometimes different results. As regards to evaluation, performance observation test were used for assessing skills and attitudes. If teachers use well-prepared Likert scale, evaluation of all groups within a reasonablely short period of time will be possible. The most effective and successful programme in therms of students' participation and enjoyment, was the 'ah-nah-bah-dah-market' program, which is kind of game of the flea market. For better organized program of this kind, however, are essential, In this program, students appraise their own attitudes and behavior by responding to a written questionnaire. In addition, students were asked to record any anecdotes relating to self-appraisal of changes on one's own attitudes and behaviours. Even after the lesson, students keep recording those changes on letters to herself. Role-playing and simulation game programme was applied to a case of 'NIMBY', in which students should decide where to located a refuse dumping ground. For this kind of programme to e successful, concepts and words used in the script should be appropriate for students' intellectual levels, and students should by adequately introduced into the objective and the procedures of the lessons. Written questionnaire was used to assess individual students' attitudes after the lesson, but in order to acquire information on the changes of students' attitudes and skills, pre-test may have to be made. Doing inquiry programme, in which advantages in which students actually investigated the environmental influence of the areas where school os located, had advantages in developing students' ability to study the environmental problems and to present the results of their studies. For this programme to be more efficient, areas of investigation should be clearly divided and alloted to each group so that repetition or overlap in areas of study and presentation be avoided, and complementary wok between groups bee enhanced. In this programme, teacher assessed students' knowledge and attitudes on the basis of reports prepared by each group. However, there were found some difficults in assessing students' attitudes and behaviours solely on the grounds of written report. Perhaps, using a scaled checklist assessing students' attitudes while their presentation could help to relieve the difficulties.

• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.351-360
• /
• 2018

Human emotion recognition is a research topic that is receiving continuous attention in computer vision and artificial intelligence domains. This paper proposes a method for classifying human emotions through multiple neural networks based on multi-modal signals which consist of image, landmark, and audio in a wild environment. The proposed method has the following features. First, the learning performance of the image-based network is greatly improved by employing both multi-task learning and semi-supervised learning using the spatio-temporal characteristic of videos. Second, a model for converting 1-dimensional (1D) landmark information of face into two-dimensional (2D) images, is newly proposed, and a CNN-LSTM network based on the model is proposed for better emotion recognition. Third, based on an observation that audio signals are often very effective for specific emotions, we propose an audio deep learning mechanism robust to the specific emotions. Finally, so-called emotion adaptive fusion is applied to enable synergy of multiple networks. The proposed network improves emotion classification performance by appropriately integrating existing supervised learning and semi-supervised learning networks. In the fifth attempt on the given test set in the EmotiW2017 challenge, the proposed method achieved a classification accuracy of 57.12%.