• The Journal of Engineering Research
• /
• v.4 no.1
• /
• pp.47-54
• /
• 2002

We consider the problem of determining smoothing parameters of Gibbs priors for Bayesian methods used in the medical imaging application of emission tomographic reconstruction. We address a simple smoothing prior (membrane) whose global hyperparameter (the smoothing parameter) controls the bias/variance tradeoff of the solution. We base our maximum-likelihood(ML) estimates of hyperparameters on observed training data, and argue the motivation for this approach. Good results are obtained with a simple ML estimate of the smoothing parameter for the membrane prior.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.06a
• /
• pp.225-226
• /
• 2018

홀로그램 콘텐츠의 정보보안을 위한 암호화 방법을 제안한다. 제안하는 암호화 기법은 실시간 처리를 위해 부분 홀로그램에 대하여 편향치를 더하고, 이중 프레넬 변환을 수행하여 에너지가 집중된 DC 영역을 획득한다. 이때 집중된 영역이 암호화 영역으로 적은 데이터롤 이용하여 고효율의 암호화를 수행한다. 제안한 기법은 기존 연구보다 변환하는 크기를 줄이기 때문에 같은 효율로 고속의 암호화를 수행할 수 있다. $1,024{\times}1,024$ 크기의 홀로그램을 $32{\times}32$ 부분홀로그램으로 구성하여 적용할 경우 약 18배 빠르게 처리할 수 있다.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.4
• /
• pp.9-17
• /
• 2022

The state-of-the-art AI systems pose many ethical issues ranging from massive data collection to bias in algorithms. In response, this paper proposes a more ethic-friendly AI architecture by combining Federated Learning(FL) and Blockchain. We discuss the importance of each issues and provide requirements for an ethical AI system to show how our solutions can achieve more ethical paradigms. By committing to our design, adopters can perform AI services more ethically.

• Annual Conference of KIPS
• /
• 2022.11a
• /
• pp.517-519
• /
• 2022

뉴럴 네트워크는 학습에 사용하는 파라미터를 문제에 맞게 최적화하여 일반화 성능을 향상시키는 것이 목적이다. 선행 연구들은 다차원의 로스 랜드스케이프(loss landscape)를 시각화하는 방법을 탐구하며, 모델의 일반화 측면에서 어떤 영향을 주는지 탐구한다. 하지만 아직까지 로스 랜드스케이프가 근본적으로 일반화 성능에 어떠한 영향을 주는지 잘 알려져 있지 않으며, 평평하거나 경사진 로스 랜드스케이프 중 어떤 형태가 일반화 성능에 더 효과적인지 여러 의견이 나뉜다. 따라서 우리는 로스 랜드스케이프가 일반화 성능과 연관 있음을 실험을 통해 파악한다. 나아가 비전문제에서 MSA(multi-head self-attention) 레이어를 기반으로 구성된 트랜스포머 구조를 사용해 작은 유도 편향(inductive bias)을 가지며 소규모 데이터 셋 체제에서의 단점을 보완한다. 결론적으로 평평한 로스 랜드스케이프가 일반화 성능에 긍정적인 영향을 끼친다는 것을 관찰한다.

• 한국벤처창업학회:학술대회논문집
• /
• 2021.11a
• /
• pp.57-61
• /
• 2021

'증거기반 교육'은 개인적 경험이나 성공 사례, 그리고 전통적인 속설보다는 과학적 연구결과와 근거가 중심이 되는 교육이다. 증거기반 창업 교육은 기존 속설과 믿음, 단편적 성공 사례로 인해 고착된 인지 편향을 완화시켜 중립적인 시각을 견지할 수 있으며, 직관과 경험을 넘어 데이터와 연구 결과에 의해 의사결정을 하는 분석적 자질을 연마하는데 기여한다. 본 논문은 현재 국내의 증기기반 창업교육의 현주소를 명확히 파악하기 위하여 1999년부터 2021년 출판된 49권의 창업교육 대학 교재를 분석하였다. 구체적으로 모든 도서의 내용을 1)창업기초, 2)비즈니스모델, 3)마케팅계획, 4)재무계획, 5)운영계획, 6)창업유형, 7)절차 및 제도의 각 7가지 기준으로 구분하고, 각각 사례, 단순통계, 변인 통계, 선행 연구의 비중을 분석하였다. 분석결과 증거의 핵심인 선행연구의 비중은 전체 교재의 총 분량 중 11.25%를 차지하는 것으로 나타났다. 이는 Charlier(2011)의 MBA 교과목 대상으로 조사한 결과와 유사한 값이다.

• Journal of the Korean Data and Information Science Society
• /
• v.18 no.2
• /
• pp.327-344
• /
• 2007

Ridge partial least squares regression (RPLS) is a regression method which can be obtained by combining ridge regression and partial least squares regression and is intended to provide better predictive ability and less sensitive to overfitting. In this paper, explicit expressions for the shrinkage factor of RPLS are developed. The structure of the shrinkage factor is explored and compared with those of other biased regression methods, such as ridge regression, principal component regression, ridge principal component regression, and partial least squares regression using a near infrared data set.

• Smart Media Journal
• /
• v.12 no.2
• /
• pp.9-14
• /
• 2023

In this paper, we proposed a face recognition network which attempts to use more facial features awhile using smaller number of training sets. When combining the neural network together for face recognition, we want to use networks that use different part of the facial features. However, the network training chooses randomly where these facial features are obtained. Other hand, the judgment basis of the network model can be expressed as a saliency map through gradCAM. Therefore, in this paper, we use gradCAM to visualize where the trained face recognition model has made a observations and recognition judgments. Thus, the network combination can be constructed based on the different facial features used. Using this approach, we trained a network for small face recognition problem. In an simple toy face recognition example, the recognition network used in this paper improves the accuracy by 1.79% and reduces the equal error rate (EER) by 0.01788 compared to the conventional approach.

• The Journal of Society for e-Business Studies
• /
• v.18 no.4
• /
• pp.315-326
• /
• 2013

Users in social media post various types of opinions such as product reviews and movie reviews. It is a common trend that customers get assistance from the opinions in making their decisions. However, as opinion usage grows, distorted feedbacks also have increased. For example, exaggerated positive opinions are posted for promoting target products. So are negative opinions which are far from common evaluations. Finding these biased opinions becomes important to keep social media reliable. Techniques of opinion mining (or sentiment analysis) have been developed to determine sentiment polarity of opinionated documents. These techniques can be utilized for finding the biased opinions. However, the previous techniques have some drawback. They categorize the text into only positive and negative, and they also need a large amount of training data to build the classifier. In this paper, we propose methods for discovering the biased opinions which are skewed from the overall common opinions. The methods are based on angle based outlier detection and personalized PageRank, which can be applied without training data. We analyze the performance of the proposed techniques by presenting experimental results on a movie review dataset.

• Korean Journal of Optics and Photonics
• /
• v.19 no.3
• /
• pp.182-186
• /
• 2008

The prism spectrometer is a standard device for the measurement of refractive index; it is used in undergraduate laboratories. Typically, however, lots of attention is required in the alignment, and the accuracy of the obtained refractive index is not so high in spite of the durability of the device. The maximum and minimum deviation method, which compensates the disadvantages of the prism spectrometer, can be composed cost effectively using a length marking tape and a rotating platform. It can measure the refractive indices accurately by utilizing a wide screen. In this study, the equal sided hollow prism whose length is 26 mm was fabricated and measured the refractive indices of seven kind of liquids (pure water, $C_3H_5(OH)_2$, $CCl_4$, $C_6H_4NH_2$, $CS_2$, $C_6H_4(CH_3)_2)$ by using the prism spectrometer and maximum and minimum deviated laser beam method at the wavelengths of He-Ne laser (${\lambda}$= 632.8 nm) and YVO4 laser (${\lambda}$= 532 nm). The result shows that the data obtained by the latter method are more accurate and precise than those obtained by the former device.

• Journal of Digital Convergence
• /
• v.19 no.5
• /
• pp.175-186
• /
• 2021

The roles and functions of domestic public libraries are diversifying, but various problems have emerged due to internally biased book lending. In addition, due to the 4th Industrial Revolution, public libraries have introduced a book recommendation system focusing on popular books, but the variety of books that users can access is limited. Therefore, in this study, the public library unborrowed book recommendation system was implemented limiting its spatial scope to Duryu Library in Daegu City to enhance the satisfaction of public library users, by using the loan records data (213,093 cases), user information (35,561 people), etc. and utilizing methods like cluster analysis, topic modeling, content-based filtering recommendation algorithm, and conducted a survey on actual users' satisfaction to present the possibility and implications of the unborrowed book recommendation system. As a result of the analysis, the majority of users responded with high satisfaction, and was able to find the satisfaction was relatively high in the class classified by specific gender, age, occupation, and usual reading. Through the results of this study, it is expected that some problems such as biased book lending and reduced operational efficiency of public libraries can be improved, and limitations of the study was also presented.