• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2023년도 추계학술발표대회
• /
• pp.406-409
• /
• 2023

With the explosion of information, recommendation algorithms are becoming increasingly important in providing people with appropriate content, enhancing their online experience. In this paper, we propose a recommender system using advanced deep reinforcement learning(DRL) techniques. This method is more adaptive and integrative than traditional methods. We selected the MovieLens dataset and employed the precision metric to assess the effectiveness of our algorithm. The result of our implementation outperforms other baseline techniques, delivering better results for Top-N item recommendations.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권3호
• /
• pp.881-895
• /
• 2023

Modern image inpainting techniques based on deep learning have achieved remarkable performance, and more and more people are working on repairing more complex and larger missing areas, although this is still challenging, especially for facial image inpainting. For a face image with a huge missing area, there are very few valid pixels available; however, people have an ability to imagine the complete picture in their mind according to their subjective will. It is important to simulate this capability while maintaining the identity features of the face as much as possible. To achieve this goal, we propose a three-stage network model, which we refer to as the identity and structure feature refinement network (ISFRNet). ISFRNet is based on 1) a pre-trained pSp-styleGAN model that generates an extremely realistic face image with rich structural features; 2) a shallow structured network with a small receptive field; and 3) a modified U-net with two encoders and a decoder, which has a large receptive field. We choose structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), L1 Loss and learned perceptual image patch similarity (LPIPS) to evaluate our model. When the missing region is 20%-40%, the above four metric scores of our model are 28.12, 0.942, 0.015 and 0.090, respectively. When the lost area is between 40% and 60%, the metric scores are 23.31, 0.840, 0.053 and 0.177, respectively. Our inpainting network not only guarantees excellent face identity feature recovery but also exhibits state-of-the-art performance compared to other multi-stage refinement models.

• 한국컴퓨터정보학회논문지
• /
• 제19권9호
• /
• pp.197-204
• /
• 2014

AS(Autonomous System) 안에서 라우터들끼리 라우팅 정보를 주고 받기 위한 내부용 라우팅 프로토콜(IGP : Interior Gateway Protocol)인 RIP, EIGRP, OSPF에 대하여 metric을 이용한 경로선택 방법들이 연구되고 있으나, 학습자들이 이론으로 이해한 내용을 실습을 통해 검증하는 구현 사례는 많지 않았다. 각 라우팅 프로토콜별로 해당 토폴로지에 기반 한 Cost value를 이론적으로 계산하여 Best Path를 결정한 후, 시뮬레이터 상에서 가상망을 구현하여 각 Routing Protocol 별로 경로선택을 검증한 결과와 서로 일치함을 확인하였다. 본 논문에서 제안한 학습방안을 활용하면 라우팅 프로토콜의 경로선택 과정을 체계적으로 이해할 수 있어 우수한 학습 결과를 성취할 수 있을 것으로 기대된다.

• 경영과학
• /
• 제30권1호
• /
• pp.103-124
• /
• 2013

This study aims to develop a framework to measure MOT competitiveness of enterprises while proposing a concept called management of technology competitiveness (MOTC). The framework of MOTC based on both resource-based view and competence-based view is consisted of technology competitiveness and management competitiveness. A variety of metrics to measure MOTC are extracted through substantial literature review. As technology competitiveness metric, this study examines R&D investment, R&D workforce, R&D facilities, intellectual property assets, and utilization of information and communication technology; as metric of management competitiveness, leadership competitiveness, maturity of the R&D systems, collaboration and partnership, learning and innovation, and commercialization are considered. We then confirm and derive the multi-dimensions of MOTC through its reliability and validity analysis. The study is expected to provide useful guidelines and references for enterprises' self-evaluation of technology and management competitiveness that is equally applicable to small, medium, and large enterprises that must compete in the global marketplace.

• 한국융합학회논문지
• /
• 제13권3호
• /
• pp.55-65
• /
• 2022

인터넷 기술과 모바일 등의 기기 발전으로 교육환경도 전통적이고 수동적인 방식에서 학습자 중심의 능동적인 방식으로 변화하고 있다. 이에 따라 학습자 개개인의 수준별 맞춤 교육의 역할도 커지고 있으며, 이에는 개별 학습자의 프로파일 구축이 중요하다. 기존의 ICT 기반 맞춤형 영어 교육의 다수는 어휘에 초점을 맞추고 있으며, 학습 콘텐츠에 대한 분석에 많은 노력을 기울이고 있다. 본 논문에서는 보다 정밀하게 사용자의 학습상태를 정의하기 위하여 단어와 문법을 대상으로 학습 상태를 구축하였다. 그리고 학습자가 특정 콘텐츠에 얼마나 익숙한지를 알려주는 콘텐츠에 대한 숙련도 메트릭을 정의하였다. 이후 실제 영문 에세이 데이터를 기반으로 사전학습을 통하여 사용자들의 숙련도를 결정하고, 시뮬레이션을 통하여 평가 에세이 데이터에 대하여 적용성이 있음을 보였다. 또한 본 연구에서 제안한 분석기법은 학습상황에 대하여 통계치나 그래프를 제공하고 학습자 수준에 적합한 학습자료를 생성하는데 필요한 데이터를 제공할 수 있다.

• 한국멀티미디어학회논문지
• /
• 제21권10호
• /
• pp.1182-1194
• /
• 2018

The Dynamic Adaptive Streaming over HTTP(DASH) is envisioned to evolve to meet an increasing demand on providing seamless video streaming services in the near future. The DASH performance heavily depends on the client's adaptive quality selection algorithm that is not included in the standard. The existing conventional algorithms are basically based on a procedural algorithm that is not easy to capture and reflect all variations of dynamic network and traffic conditions in a variety of network environments. To solve this problem, this paper proposes a novel quality selection mechanism based on the Deep Q-Network(DQN) model, the DQN-based DASH Adaptive Bitrate(ABR) mechanism. The proposed mechanism adopts a new reward calculation method based on five major performance metrics to reflect the current conditions of networks and devices in real time. In addition, the size of the consecutive video segment to be downloaded is also considered as a major learning metric to reflect a variety of video encodings. Experimental results show that the proposed mechanism quickly selects a suitable video quality even in high error rate environments, significantly reducing frequency of quality changes compared to the existing algorithm and simultaneously improving average video quality during video playback.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권4호
• /
• pp.2020-2037
• /
• 2017

In order to improve the energy efficiency of n-Epidemic routing protocol in opportunistic networks, in which a stable end-to-end forwarding path usually does not exist, a novel adjustment strategy for parameter n is proposed using learning atuomata principle. First, nodes dynamically update the average energy level of current environment while moving around. Second, nodes with lower energy level relative to their neighbors take larger n avoiding energy consumption during message replications and vice versa. Third, nodes will only replicate messages to their neighbors when the number of neighbors reaches or exceeds the threshold n. Thus the number of message transmissions is reduced and energy is conserved accordingly. The simulation results show that, n-Epidemic routing protocol with the proposed adjustment method can efficiently reduce and balance energy consumption. Furthermore, the key metric of delivery ratio is improved compared with the original n-Epidemic routing protocol. Obviously the proposed scheme prolongs the network life time because of the equilibrium of energy consumption among nodes.

• Journal of Electrical Engineering and Technology
• /
• 제9권5호
• /
• pp.1739-1745
• /
• 2014

Prediction of fault-prone modules continues to attract researcher's interest due to its significant impact on software development cost. The most important goal of such techniques is to correctly identify the modules where faults are most likely to present in early phases of software development lifecycle. Various software metrics related to modules level fault data have been successfully used for prediction of fault-prone modules. Goal of this research is to predict the faulty modules at design phase using design metrics of modules and faults related to modules. We have analyzed the effect of pre-processing and different machine learning schemes on eleven projects from NASA Metrics Data Program which offers design metrics and its related faults. Using seven machine learning and four preprocessing techniques we confirmed that models built from design metrics are surprisingly good at fault proneness prediction. The result shows that we should choose Naïve Bayes or Voting feature intervals with discretization for different data sets as they outperformed out of 28 schemes. Naive Bayes and Voting feature intervals has performed AUC > 0.7 on average of eleven projects. Our proposed framework is effective and can predict an acceptable level of fault at design phases.

• International Journal of Computer Science & Network Security
• /
• 제22권1호
• /
• pp.225-233
• /
• 2022

The recent outbreak of corona virus (COVID-19) infectious disease had made its forecasting critical cornerstones in most scientific studies. This study adopts a machine learning based time series model - Auto Regressive Integrated Moving Average (ARIMA) model to forecast COVID-19 confirmed cases in Botswana over 60 days period. Findings of the study show that COVID-19 confirmed cases in Botswana are steadily rising in a steep upward trend with random fluctuations. This trend can also be described effectively using an additive model when scrutinized in Seasonal Trend Decomposition method by Loess. In selecting the best fit ARIMA model, a Grid Search Algorithm was developed with python language and was used to optimize an Akaike Information Criterion (AIC) metric. The best fit ARIMA model was determined at ARIMA (5, 1, 1), which depicted the least AIC score of 3885.091. Results of the study proved that ARIMA model can be useful in generating reliable and volatile forecasts that can used to guide on understanding of the future spread of infectious diseases or pandemics. Most significantly, findings of the study are expected to raise social awareness to disease monitoring institutions and government regulatory bodies where it can be used to support strategic health decisions and initiate policy improvement for better management of the COVID-19 pandemic.

• 한국방사선학회논문지
• /
• 제14권6호
• /
• pp.773-780
• /
• 2020

4차 산업의 발전으로 의학·보건·바이오 등 여러 과학기술 분야에서는 질병을 예방하고 질병에 대한 피해를 줄이기 위한 연구가 이루어지고 있으며, 최근에는 ICT 기술의 발전과 더불어 인공지능 기술이 급부상하고 그 효용성이 입증되면서 영상의학 검사의 영상 분석에 인공지능 기술이 도입되어 연구되고 있다. 본 논문에서는 흉부 X선 영상을 이용하여 폐렴의 분류와 검출에 대한 딥러닝 모델을 직접 적용해보고 실제로 Inception 계열의 딥러닝 모델이 폐렴 검출에 있어 유용한 모델인지 평가하고자 한다. 실험재료는 캐글(Kaggle)에서 무료로 제공 및 공유하는 흉부 X선 영상 데이터 세트를 사용하였으며 전체 3,470개의 흉부 X선 영상 데이터 중 학습 데이터 세트 1,870개, 검증 데이터 세트 1,100개, 테스트 데이터 세트 500개로 분류하였다. 실험결과 Inception V3 딥러닝 모델의 Metric 평가에 대한 결과값은 정확도는 94.80%, 정밀도는 97.24%, 재현율은 94.00%, F1 스코어는 95.59의 결과값을 나타내었다. 그리고 흉부 X선 영상의 페렴 검출 및 분류에 대하여 Inception V3 딥러닝 모델링에 대한 최종 에포크의 정확도는 학습 모델링의 경우 94.91%, 검증 모델링은 89.68%의 정확도를 나타내었다. 손실함수 값의 평가는 학습 모델링은 1.127%, 검증 모델링은 4.603%의 손실함수 값을 나타내었다. 이러한 결과로 Inception V3 딥러닝 모델은 흉부영상 데이터의 특징 추출 및 분류에 있어 매우 우수한 딥러닝 모델이며 학습상태 또한 매우 우수하다고 평가하였다. 테스트 모델링에 대한 매트릭스 정확도 평가 결과 정상 흉부 X선 영상 데이터의 경우 96%, 폐렴 흉부 X선 영상데이터의 경우 97%의 정확도가 입증되었다. Inception 계열의 딥러닝 모델의 경우 흉부 질환의 분류에 있어 유용한 딥러닝 모델이 될 것이라고 판단되며 인력의 보조적인 역할 또한 수행할 수 있을 것이라고 기대되어 부족한 의료인력 문제에도 해결점이 될 것이라고 사료된다. 향후 딥러닝을 이용한 폐렴의 진단에 대한 유사 연구 시 본 연구는 유사 연구의 기초자료로 제시될 것이라고 기대된다.