• 한국멀티미디어학회논문지
• /
• 제25권10호
• /
• 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%.

• International Journal of Computer Science & Network Security
• /
• 제22권6호
• /
• pp.45-50
• /
• 2022

The article substantiates the feasibility of using and actively implementing innovative technologies in the practice of organizing the educational process. The need for the use of telecommunication technologies, which provide constant communication between students and the teacher outside the classroom, has been identified. Particular attention is paid to the latest approaches to the use of various forms of multimedia technologies in student education, which intensify the process of acceptance and assimilation of educational material by foreign students. The advantages of using innovative means of distance education are determined, which thanks to modern electronic educational systems allow students to receive quality higher education. Innovative technologies promote the development of cognitive interest in students, they learn to systematize and summarize the material studied, discuss and debate. In this regard, the reorientation of the system of higher education in Europe towards innovation is becoming the most important tool in ensuring the competitiveness of graduates in the labor market. In addition, the investment attractiveness of a university often depends on the innovative nature of the development of scientific, educational and practical activities of the subjects of the educational process, their inclusion in the national innovation system. The article analyzes that in the universities of the European Union in the training of specialists in the management of basic interactive methods, forms and tools are binary lecture, briefing, webinar, video conference, video lecture, virtual consultation, virtual tutorial, slide lecture, comp. utheric tests. Various classes on slide technology took active forms during the training of management specialists.

• 한국컴퓨터정보학회논문지
• /
• 제27권2호
• /
• pp.15-23
• /
• 2022

특허 분야에서 자연어처리(Natural Language Processing) 태스크는 특허문헌의 언어적 특이성으로 문제 해결의 난이도가 높은 과제임에 따라 한국 특허문헌에 최적화된 언어모델의 연구가 시급한 실정이다. 최근 자연어처리 분야에서는 특정 도메인에 특화되게 사전 학습(Pre-trained)한 언어모델을 구축하여 관련 분야의 다양한 태스크에서 성능을 향상시키려는 시도가 지속적으로 이루어지고 있다. 그 중, ELECTRA는 Google이 BERT 이후에 RTD(Replaced Token Detection)라는 새로운 방식을 제안하며 학습 효율성을 높인 사전학습 언어모델이다. 본 연구에서는 대량의 한국 특허문헌 데이터를 사전 학습한 KorPatELECTRA를 제안한다. 또한, 특허 문헌의 특성에 맞게 학습 코퍼스를 정제하고 특허 사용자 사전 및 전용 토크나이저를 적용하여 최적화된 사전 학습을 진행하였다. KorPatELECTRA의 성능 확인을 위해 실제 특허데이터를 활용한 NER(Named Entity Recognition), MRC(Machine Reading Comprehension), 특허문서 분류 태스크를 실험하였고 비교 대상인 범용 모델에 비해 3가지 태스크 모두에서 가장 우수한 성능을 확인하였다.

• International Journal of Computer Science & Network Security
• /
• 제22권12호
• /
• pp.91-97
• /
• 2022

Analyzing the psychological and pedagogical literature, we showed the interest of researchers in the problem posed. The concept of competence is considered, which is interpreted as giving the key to solving a wide range of educational and life tasks. Research competence implies the ability to cooperate, enter into contacts, readiness for changes, for self-determination and is an integral quality of the individual, expressed in the readiness and ability to independently search for solutions to new problems and creative transformation of reality based on a set of personal and meaningful knowledge, skills, methods of activity and value attitudes.The article offers conditions that certify the improvement of forms and methods of training students in the formation of research competence of future specialists. The use of innovative technologies contributes to improving the level of training of future specialists: students are better prepared for classes, take an active part in the assimilation of program material in laboratory classes. It is noted that this creates a subject-subject relationship between the student and the teacher, and changes the attitude of students to classes. In the process of such organization of educational activities, students are convinced of the need for knowledge and its effectiveness, learn to compare, generalize, classify, establish cause-and-effect relationships, express opinions, defend their point of view, they ensure success in their studies, and develop research competence. It is proved that in order to apply the latest technologies, the teacher himself must know them well, that is, constantly improve himself, master new methods, techniques, ideas, which will help him create new pedagogical technologies and implement them in the educational process.

• International Journal of Computer Science & Network Security
• /
• 제23권12호
• /
• pp.167-174
• /
• 2023

Electronic educational environments in the conditions of quarantine restrictions of COVID-19 have become a common phenomenon for the organization of distance educational activities. Under the conditions of Russian aggression, Ukrainian proof of their use is unique. The purpose of the article is to analyze the role of electronic educational environments in the process of training applicants for higher education in Ukraine in the realities of a large-scale war. General scientific methods (analysis, synthesis, deduction, and induction) and special pedagogical prognostic methods, modeling, and SWOT analysis methods were used. In the results, the general properties of the Internet educational platforms common in Ukraine, the peculiarities of using the Moodle and Prometheus platforms, and an approximate model of the electronic learning environment were discussed. The reasons for the popularity of Moodle among Ukrainian universities are analyzed, but vulnerable elements related to security are emphasized. It was also determined that the high cost of Prometheus software and less functionality made this learning environment less relevant. The conclusions state that the military actions drew the attention of universities in Ukraine to the formation of their own educational platforms. This is especially relevant for technical and military institutions of higher education.

• 실천공학교육논문지
• /
• 제16권3_spc호
• /
• pp.291-296
• /
• 2024

본 연구는 장애인 직업훈련생의 사전 평가 데이터를 활용하여 직업 훈련 후 취업 결과를 예측하는 다양한 머신러닝 모델을 분석하였다. 연구는 훈련생의 성별, 연령, 장애 유형 등을 포함하는 다양한 개인적 특성을 포함한 데이터 세트에 기반하여, 가장 적합한 머신러닝 모델들을 선별하고 활용하였다. 이러한 분석을 통해, 사전 평가 데이터만을 사용하여 장애인 훈련생의 취업률 및 직업 만족도 향상을 목적으로 한다. 결과적으로, 장애인뿐만 아니라 다양한 배경을 가진 직업훈련생들에게도 적용할 수 있는 범용적인 접근법을 제시한다. 이는 맞춤형 직업 훈련 프로그램의 개발과 구현에 중요한 기여를 할 것으로 기대되며, 궁극적으로는 더 나은 취업 결과와 직업 만족도를 달성하는 데 도움이 될 것이다.

• 융합보안논문지
• /
• 제13권5호
• /
• pp.57-66
• /
• 2013

미래의 전쟁 양상을 예측하는 것이 장차 전쟁에서 승리를 보장 받을 수 있다. 이라크 전에서 미군이 보여준 전쟁방식에서 미래의 전쟁양상을 도출해 낼 수 있다. 즉, GPS, 첩보위성, 무인정찰기를 이용한 실시간의 정보 수집에서 고도화된 유도 무기를 통한 정밀타격까지의 일련의 전투과정은 우주, 공중, 해상, 지상 전력을 네트워크화 함으로써 실시간 동시 및 통합 전투력 발휘가 전쟁의 승패를 결정하는 핵심 요소라는 것을 다시 한 번 확인할 수 있었다. NCW는 미래 작전유형이 작전에 참여하는 모든 요소들의 네트워킹에 의해 유기적으로 연동되는 네트워크 중심 전장을 의미한다. 따라서 정보수집 센서, 지휘 결심, 타격체계들이 지리적 여건에 제한되지 않고 네트워킹 되어 전투에 참여하는 지휘관에서부터 각개 병사에 이르기까지 원하는 시간에 원하는 정보를 획득할 수 있는 전쟁수행개념을 말한다. 미래형컴퓨터에 대하여 고찰하고, 특히 휴대용 컴퓨터, 키보드 없이도 손으로 자유롭게 사용할 수 있는 컴퓨터, 음성인식 컴퓨터, 이를 이용한 인공지능, 또한 이를 통합 운용할 수 있는 네트워크 체계를 이해하고 해병대의 종합적인 정보통신망 구축과 이를 전투나 훈련에 적용해 보고자 한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권10호
• /
• pp.3378-3393
• /
• 2014

We have witnessed the rapid development of information technology in recent years. One of the key phenomena is the fast, near-exponential increase of data. Consequently, most of the traditional data classification methods fail to meet the dynamic and real-time demands of today's data processing and analyzing needs--especially for continuous data streams. This paper proposes an improved incremental learning algorithm for a large-scale data stream, which is based on SVM (Support Vector Machine) and is named DS-IILS. The DS-IILS takes the load condition of the entire system and the node performance into consideration to improve efficiency. The threshold of the distance to the optimal separating hyperplane is given in the DS-IILS algorithm. The samples of the history sample set and the incremental sample set that are within the scope of the threshold are all reserved. These reserved samples are treated as the training sample set. To design a more accurate classifier, the effects of the data volumes of the history sample set and the incremental sample set are handled by weighted processing. Finally, the algorithm is implemented in a cloud computing system and is applied to study user behaviors. The results of the experiment are provided and compared with other incremental learning algorithms. The results show that the DS-IILS can improve training efficiency and guarantee relatively high classification accuracy at the same time, which is consistent with the theoretical analysis.

• 대한인간공학회지
• /
• 제34권2호
• /
• pp.179-190
• /
• 2015

Objective and Background: Due to increase in Musculoskeletal Disorders (MSDs) relating to computer use, a number of ergonomics recommendations have been proposed in order to tackle this problem. However, some of these recommendations have been conflicting. Method: This study was to survey the VDT (Visual Display Terminal) use of office workers. The subjects were 452 workers at 13 places of business and the data were collected by self-administered questionnaire. Results: As a result, prevalence of self-reported MSDs of all VDT workers was 90.2% and shoulder took up 57.0% and neck 38.3% by symptom part of body. The population of computer use of more than 6h/day was 84.5%, and 33.8% also reported using the VDT 2~3h/day without the rest time. Desktop computer users were 95.6%, and a 17-inch monitor accounted for 42.0% among the desktop users. As a result of satisfaction survey on overall computer work, 21.1% of the total respondents said satisfied, desk complaint was about 24.6%, and chair complaint was 33.4%. Despite the importance of computer environment, satisfaction was from fair to uncomfortable. Conclusion and Application: In conclusion, office workers are prone to the MSDs due to their work environment. Additionally, this study found that task was a significant effect for the majority of dependent variables, and therefore, the improvement of computer workstations work environment is urgent, and the improvement of desk height adjustment, chair seat size (length, width), backrest condition, location of keyboard (mouse) and arm rest is required.

• 한국컴퓨터정보학회논문지
• /
• 제17권3호
• /
• pp.67-76
• /
• 2012

트래픽 분류는 트래픽 관리하는데 중요한 역할을 차지하고 있다. 전통적인 방법은 P2P와 암호화 트래픽을 제대로 분류할 수 없는 문제가 있다. 서포트 벡터 머신은 기존의 문제를 해결할 수 있고 병목 현상을 극복할 수 있는 유용한 분류 도구이다. 하지만 서포트 벡터 머신의 주요 장점은 이차 프로그래밍(QP)문제 때문에 큰 데이터 집단을 훈련하는데 시간을 소모한다. 그러나 유용한 서포트 벡터는 전체 데이터에서 극히 일부분이다. 만약 우리가 훈련전에 쓸모없는 벡터들을 삭제할 수 있다면, 시간을 절약하고 정확도를 유지할 수 있다. 이 논문에서 우리는 대규모 데이터를 다룰 때 훈련 속도를 빠르게 하기위해 순차적인 방법을 통해 쓸모없는 벡터들을 제거하기 위한 가능성을 논의하였다.