• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권11호
• /
• pp.5496-5521
• /
• 2018

Object detection and tracking is the basic capability of mobile robots to achieve natural human-robot interaction. In this paper, an object tracking system of mobile robot is designed and validated using improved multiple instance learning algorithm. The improved multiple instance learning algorithm which prevents model drift significantly. Secondly, in order to improve the capability of classifiers, an active sample selection strategy is proposed by optimizing a bag Fisher information function instead of the bag likelihood function, which dynamically chooses most discriminative samples for classifier training. Furthermore, we integrate the co-training criterion into algorithm to update the appearance model accurately and avoid error accumulation. Finally, we evaluate our system on challenging sequences and an indoor environment in a laboratory. And the experiment results demonstrate that the proposed methods can stably and robustly track moving object.

• International Journal of Computer Science & Network Security
• /
• 제22권3호
• /
• pp.252-258
• /
• 2022

Globalization, development of the information society, intensification of migration processes, and internationalization of education in recent years have significantly affected the international labour market and increased competition between professionals in various fields. Academic, research and teaching staff are currently facing rapid changes and growing demand for quality educational services. Under such conditions, educational institutions around the world pay particular attention to improving the quality of the educational process in order for their graduates to be able to compete in the international labour market. The study of the experience of teachers' professional training in advanced countries opens new opportunities for improving the system of pedagogical training in Ukraine in order to adapt it to the requirements of the European educational space. The progressive achievements of countries demonstrating a high level of teachers' professional training in accordance with international standards, having rich historical educational traditions are of considerable scientific interest; consequently, this contributes to their leadership in science and education at the regional and global levels. The purpose of the present academic paper lies in analysing the latest trends in exploring teaching experience in the European countries and performance review of passing by students and teachers of the program on studying of pedagogical practice and experience of teaching in the countries of Europe. Methodology. Analytical and empirical (questionnaire) methods were used in the research process. Results. According to the obtained results of the research, the advantages, the most common areas of educational programs and the wishes of the participants regarding the practical implementation of the experience were identified.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2021년도 춘계학술대회
• /
• pp.2-4
• /
• 2021

분산된 환경에서 머신 러닝의 학습 가중치를 공유하여 학습하는 방법은 훈련 데이터를 직접 공유하는 것이 아니기 때문에 안전한 것으로 여겨졌다. 하지만, 최근 연구에 따르면 악의적인 공격자가 공유된 가중치를 분석하여 원본 데이터를 완벽하게 복원할 수 있는 취약점이 발견되었다. Gradient Leakage Attack은 이러한 취약점을 이용해 훈련 데이터를 복원하는 공격 기법이다. 본 연구에서는 개별 장치에서 학습을 진행하고 가중치를 서버와 공유하는 학습 환경인 연합 학습 환경에서 해당 공격을 방어하기 위해 이산 코사인 변환에 기반한 이미지 변환 기법을 제시한다. 실험 결과, 우리의 이미지 변환 기법을 적용하면 공유된 가중치로부터 원본 데이터를 완벽하게 복원할 수 없다.

• Journal of Information Processing Systems
• /
• 제20권2호
• /
• pp.149-158
• /
• 2024

Traditional manual identification of crop leaf diseases is challenging. Owing to the limitations in manpower and resources, it is challenging to explore crop diseases on a large scale. The emergence of artificial intelligence technologies, particularly the extensive application of deep learning technologies, is expected to overcome these challenges and greatly improve the accuracy and efficiency of crop disease identification. Crop leaf disease identification models have been designed and trained using large-scale training data, enabling them to predict different categories of diseases from unlabeled crop leaves. However, these models, which possess strong feature representation capabilities, require substantial training data, and there is often a shortage of such datasets in practical farming scenarios. To address this issue and improve the feature learning abilities of models, this study proposes a deep transfer learning adaptation strategy. The novel proposed method aims to transfer the weights and parameters from pre-trained models in similar large-scale training datasets, such as ImageNet. ImageNet pre-trained weights are adopted and fine-tuned with the features of crop leaf diseases to improve prediction ability. In this study, we collected 16,060 crop leaf disease images, spanning 12 categories, for training. The experimental results demonstrate that an impressive accuracy of 98% is achieved using the proposed method on the transferred ResNet-50 model, thereby confirming the effectiveness of our transfer learning approach.

• 한방재활의학과학회지
• /
• 제19권3호
• /
• pp.69-80
• /
• 2009

Objectives : The purpose of this study was to evaluate the effects of core stability training on postural control and balance of hemiplegia patients who are difficult to control posture due to stroke. Methods : Subjects of the study were consisted of 25 adult hemiplegia patients(experimental 12, control 13) who were receiving rehabilitation therapy in hospital. Its group had a core stability training program by a physical therapists for 40 minutes, five times a week for nine-week period. Measurements of postural assessment scale for stroke(PASS), berg balance scale(BBS) and manual function test(MFT) were evaluated at initial presentation(pretest) and after completion of the each therapy program(posttest). Independent t-test and paired t-test was utilized to detect the mean difference between the groups. Results : Firstly, the result of PASS showed that postural control was significantly increased after the completion of core stability training(p<.01). Secondly, the result of BBS showed that balance control was significantly increased after the completion of core stability training(p<.01) and general physical therapy(p<.01). Lastly, the result of MFT showed that upper extremity's motor function was significantly increased after the completion of core stability training(p<.001). Conclusions : This study showed that core stability training is an effective treatment strategy on postural control, balance and upper extremity's motor function.

• 한국컴퓨터정보학회논문지
• /
• 제17권5호
• /
• pp.147-155
• /
• 2012

미국 경찰 역사상 3세기 동안 경찰 교육도 상당히 개선되었음을 보여주고 있다. 경찰학교 탄생과 더불어 일반 국민이 경찰에 거는 기대도 커졌으며 경찰이 보여주는 능력도 성장했다. 1960년대 이후 현장 훈련 과정, 기관 내 훈련 등을 개발 운영해오고 있다. 최근 몇 년 동안 여러 지역의 경찰학교 훈련에서 대학식 접근법을 택하고 있는데 특히 COP식 경찰 업무 수행과 전문화를 이루도록 하는데 목표를 두고 있다. 우리나라는 관련법규의 미정비, 비체계적인 교육과정, 학교교육과 실습교육의 Switch System의 한계, 열악한 교육여건 및 환경 등이 문제점으로 대두되고 있어 개선을 위한 방안은 첫째, 교육훈련 법규를 통합, 정비 둘째, 교과과정을 체계적으로 개편 셋째, 교육훈련효과의 극대화를 위하여 Switch System을 정착 넷째, 교육여건 및 교육환경을 개선시켜야 한다.

• 벤처창업연구
• /
• 제13권5호
• /
• pp.55-61
• /
• 2018

본 논문은 기업의 교육훈련에 대한 투자가 기업의 성과에 영향을 미치는 변수별 효과를 살펴봄으로써 기업의 교육훈련에 대한 관심과 투자를 촉진하고, 이를 지원하기 위한 정책적 함의를 도출하는데 목적이 있다. 분석자료는 HCCP를 사용하였으며, 분석방법은 동태패널 모형과 변수의 내생성 문제해결을 위한 방법을 적용하였다. 분석 결과, 기업의 교육훈련이 1인당 매출액에는 양(+)의 영향을 미치나 재무성과인 1인당 순이익에 미치는 영향은 일부 모형에서만 양(+)의 효과로 나타났다. 그러나 내생성을 고려한 모형에서는 1인당 순이익이 유의한 양(+)의 효과로 나타나 변수의 내생성도 중요한 고려요인임을 확인하였다. 이를 통해 기업에서는 단기적인 성과보다는 중장기적인 관점에서 교육훈련에 대한 관심과 투자를 확대해 나간다면 기업의 조직성과 및 재무성과에 크게 기여할 것이다.

• International Journal of Computer Science & Network Security
• /
• 제22권4호
• /
• pp.387-393
• /
• 2022

The article reflects short-term perspectives on the use of information and communication technologies in the training of teachers for higher education. Education is characterized by conservatism, so aspects of systematic development of the industry are relevant to this cluster of social activity. Therefore, forecasting the introduction of innovative elements of ICT training is in demand for the educational environment. Forecasting educational trends are most relevant exactly in the issues of training future teachers of higher education because these specialists are actually the first to implement the acquired professional skills in pedagogical activities. The article aims to consider the existing potential of ICT-based learning, its implementation in the coming years, and promising innovative educational elements that may become relevant for the educational space in the future. The tasks of scientific exploration are to show the optimal formats of synergy between traditional and innovative models of learning. Based on already existing experience, extrapolation of conditions of educational process organization with modeling realities of using information and communication technologies in various learning dimensions should be carried out. Educational trends for the next 3 years are a rather tentative forecast because, as demonstrated by the events associated with the COVID-19 pandemic, the socio-cultural space is very changeable. Consequently, the dynamism of the educational environment dictates the need for a value-based awareness of the information society and the practical use of technological advances. Thus, information and communication technologies are a manifestation of innovative educational strategies of today and become an important component along with traditional aspects of educational process organization. Future higher education teachers should develop a training strategy taking into account the expediency of the ICT component.

• 운동영양학회지
• /
• 제24권2호
• /
• pp.11-21
• /
• 2020

[Purpose] Doxorubicin (DOX) is a potent anti-cancer drug that appears to have severe myotoxicity due to accumulation. The skeletal muscle has a regeneration capacity through satellite cell activation when exposed to extracellular stimulus or damage. Endurance exercise (EXE) is a therapeutic strategy that improves pathological features and contributes to muscle homeostasis. Thus, this study investigated the effect of EXE training in mitigating chronic DOX-induced myotoxicity. [Methods] Male C57BL/6J mice were housed and allowed to acclimatize with free access to food and water. All the mice were randomly divided into four groups: sedentary control (CON, n=9), exercise training (EXE, n=9), doxorubicin treatment (DOX, n=9), doxorubicin treatment and exercise training (DOX+EXE, n=9) groups. The animals were intraperitoneally injected with 5 mg/kg/week of DOX treatment for 4 weeks, and EXE training was initiated for treadmill adaptation for 1 week and then performed for 4 weeks. Both sides of the soleus (SOL) muscle tissues were dissected and weighed after 24 hours of the last training sessions. [Results] DOX chemotherapy induced an abnormal myofiber's phenotype and transition of myosin heavy chain (MHC) isoforms. The paired box 7 (PAX7) and myoblast determination protein 1 (MYOD) protein levels were triggered by DOX, while no alterations were shown for the myogenin (MYOG). DOX remarkably impaired the a-actinin (ACTN) protein, but the EXE training seems to repair it. DOX-induced myotoxicity stimulated the expression of the forkhead box O3 (FOXO3a) protein, which was accurately controlled and adjusted by the EXE training. However, the FOXO3a-mediated downstream markers were not associated with DOX and EXE. [Conclusion] EXE postconditioning provides protective effects against chronic DOX-induced myotoxicity, and should be recommended to alleviate cancer chemotherapy-induced late-onset myotoxicity.

• 인터넷정보학회논문지
• /
• 제25권2호
• /
• pp.1-9
• /
• 2024

연합학습은 보안 및 프라이버시 측면에서 중앙 집중식 방법보다 안전하도록 설계되었음에도 불구하고 여전히 많은 취약점을 내재한다. 적대적 공격(adversarial attack)을 수행하는 공격자는 신중하게 제작된 입력 데이터, 즉 적대적 예제(adversarial examples)를 클라이언트의 학습 데이터에 주입하여 딥러닝 모델을 의도적으로 조작하여 오분류를 유도한다. 이에 대한 보편적인 방어 전략은 이른바 적대적 학습(adversarial training)으로 적대적 예제들의 특성을 선제적으로 모델에 학습시키는 것이다. 기존의 연구에서는 모든 클라이언트가 적대적 공격 하에 있는 상황을 가정하는데 연합학습의 클라이언트 수가 매우 많음을 고려하면 실제와는 거리가 있다. 본 논문에서는 클라이언트의 일부가 공격 하에 있는 시나리오에서 적대적 학습의 양상을 실험적으로 살핀다. 우리는 실험을 통해 적대적 예제에 대한 분류 정확도가 증가하면 정상 샘플에 대한 분류 정확도의 감소하는 트레이드오프 관계를 가짐을 밝혔다. 이러한 트레이드오프 관계를 효과적으로 활용하기 위해 클라이언트가 자신이 공격받는지 여부에 따라 손실함수를 적응적으로 선택하여 적대적 학습을 수행하는 방법을 제시한다.