• Journal of Information Processing Systems
• /
• 제14권6호
• /
• pp.1464-1479
• /
• 2018

In face recognition, sometimes the number of available training samples for single category is insufficient. Therefore, the performances of models trained by convolutional neural network are not ideal. The small sample face recognition algorithm based on novel Siamese network is proposed in this paper, which doesn't need rich samples for training. The algorithm designs and realizes a new Siamese network model, SiameseFacel, which uses pairs of face images as inputs and maps them to target space so that the $L_2$ norm distance in target space can represent the semantic distance in input space. The mapping is represented by the neural network in supervised learning. Moreover, a more lightweight Siamese network model, SiameseFace2, is designed to reduce the network parameters without losing accuracy. We also present a new method to generate training data and expand the number of training samples for single category in AR and labeled faces in the wild (LFW) datasets, which improves the recognition accuracy of the models. Four loss functions are adopted to carry out experiments on AR and LFW datasets. The results show that the contrastive loss function combined with new Siamese network model in this paper can effectively improve the accuracy of face recognition.

• Biomolecules & Therapeutics
• /
• 제27권2호
• /
• pp.231-239
• /
• 2019

Suppressor of Variegation 3-9 Homolog 2 (SUV39H2) methylates the lysine 9 residue of histone H3 and induces heterochromatin formation, resulting in transcriptional repression or silencing of target genes. SUV39H1 and SUV39H2 have a role in embryonic development, and SUV39H1 was shown to suppress cell cycle progression associated with Rb. However, the function of human SUV39H2 has not been extensively studied. We observed that forced expression of SUV39H2 decreased cell proliferation by inducing $G_1$ cell cycle arrest. In addition, SUV39H2 was degraded through the ubiquitin-proteasomal pathway. Using yeast two-hybrid screening to address the degradation mechanism and function of SUV39H2, we identified translationally controlled tumor protein (TCTP) as an SUV39H2-interacting molecule. Mapping of the interacting regions indicated that the N-terminal 60 amino acids (aa) of full-length SUV39H2 and the C-terminus of TCTP (120-172 aa) were critical for binding. The interaction of SUV39H2 and TCTP was further confirmed by co-immunoprecipitation and immunofluorescence staining for colocalization. Moreover, depletion of TCTP by RNAi led to up-regulation of SUV39H2 protein, while TCTP overexpression reduced SUV39H2 protein level. The half-life of SUV39H2 protein was significantly extended upon TCTP depletion. These results clearly indicate that TCTP negatively regulates the expression of SUV39H2 post-translationally. Furthermore, SUV39H2 induced apoptotic cell death in TCTP-knockdown cells. Taken together, we identified SUV39H2, as a novel target protein of TCTP and demonstrated that SUV39H2 regulates cell proliferation of lung cancer cells.

• 사물인터넷융복합논문지
• /
• 제8권5호
• /
• pp.77-83
• /
• 2022

낸드 플래시 메모리는 구조적으로 쓰기 전 지우기(Erase-Before-Write) 동작이 요구된다. 이것을 해결하기 위해서는 데이터 업데이트 동작이 빈번히 발생하는 페이지(Hot data page)를 구분하여 별도에 블록에 저장함으로 해결할 수 있으며 이러한 Hot data를 분류하는 기법을 핫 데이터 판단기법이라 한다. MHF(Multi Hash Function Framework)기법은 데이터 갱신요청의 빈도를 시스템 메모리에 기록하고 그 기록된 값이 일정 기준 이상일 때 해당 데이터 갱신요청을 Hot data로 판단한다. 하지만 데이터 갱신요청에 빈도만을 단순히 카운트하는 방법으로는 정확한 Hot data로 판단에 한계가 있다. 또한 데이터 갱신요청의 지속성을 판단 기준으로 하는 기법의 경우 갱신요청 사실을 시간 간격을 기준으로 순차적으로 기록한 뒤 Hot data로 판단하는 방법이다. 이러한 지속성을 기준으로 하는 방법의 경우 그 구현과 운용이 복잡한 단점이 있으며 갱신요청에 빈도를 고려하지 않는 경우 부정확하게 판단되는 문제가 있다. 본 논문은 데이터 갱신요청에 빈도와 지속성을 함께 고려한 경량화된 핫 데이터 판단기법을 제안한다.

• 한국음향학회지
• /
• 제26권3호
• /
• pp.123-128
• /
• 2007

본 논문에서는 수중 천이 신호에 대한 식별 알고리즘을 제안한다. 일반적으로 해양의 배경잡음은 스펙트럼 특성 및 에너지 변화가 적은 정재성을 갖는 반면에 천이 신호는 스펙트럼 및 에너지 변화가 큰 비정재성을 가진다. 따라서 수중 천이 신호 식별을 위하여 선행되어져야 하는 수중 천이 신호 탐지에서는 프레임 단위로 스펙트럼 변이와 에너지 변화를 이용한다. 제안한 수중 천이 신호 식별 알고리즘에서는 특징 벡터를 추출하기 위하여 위그너-빌 분포 함수를 기반으로 고유치 분해를 이용한다. 추출된 특징 벡터를 기반으로 탐지된 수중 천이 신호의 특징 벡터와 식별하고자 하는 데이터베이스에 있는 기준 신호의 특징 벡터와의 상관 값을 프레임 단위로 계산하고, 각 클래스별로 프레임 사상도를 산출하여 최대 값을 갖는 기준 신호로 탐지된 수중 천이 신호를 식별한다.

• 한국음향학회지
• /
• 제18권7호
• /
• pp.101-111
• /
• 1999

본 논문에서는 정확한 콘크리트 설계강도 분류를 위해 인공지능 기법에 바탕을 둔 증거축적방법에 의한 초음파신호의 패턴인식방법을 제안하였다. 이를 위해 우선 초음파신호의 특징파라메터로 분산, 영점교차횟수, 평균주파수, 자기회귀모델계수 및 선형 켑스트럼계수를 추출하였다. 추출된 특징파라메터들의 각각의 특성을 알아보고, 하나의 특징파라메터로 설계강도의 정확한 분류가 어렵다는 것을 보였다. 이러한 문제점을 해결하기 위하여 추출된 다수의 특징파라메터들을 이용하여 설계강도 분류를 증거축적방법을 통해 수행하였다. 또한, 이 증거축적방법을 콘크리트 패턴인식에 적용하기 위해 퍼지매핑 함수를 도입하였다. 본 논문에서 제안한 알고리즘이 다수의 특징파라메터들을 효율적으로 이용하여 92%의 패턴인식률을 보였으며, 이는 기존의 패턴 분류 알고리즘보다 콘크리트 설계강도를 보다 정확하게 분류함을 확인하였다.

• 전기학회논문지
• /
• 제59권4호
• /
• pp.809-813
• /
• 2010

Although the goals of HL7 Version 2(V2) and Version 3(V3) are identical, the concepts of the implementation and technological basis are different; this makes their versions inconvertible. This problem interrupts technological innovation advanced from V2 to V3 and has been raised as a new type of barrier in the field of medical information system. This study intends to develop software to convert V2 to V3 which can be utilized in the actual medical environment. Since it is practically difficult to develop the whole tools that automatically change total V2 messages into V3 messages, this article has designed, implemented, and tested the software that allows mapping between V2 and V3 which function as tools. In order to test this program, it has used ADT^A03 of five V2 messages to generate V3 messages. It is expected that the result of this research will be one of the new methods allowing conversion between V2 and V3.

• 대한조선학회지
• /
• 제14권3호
• /
• pp.13-22
• /
• 1977

The hydrodynamic forces acting on a forced oscillating 2-dimensional cylinder on a free surface of a fluid of a finite depth are calculated by distributing singularities on the immersed body surface. And the Haskind-Newman relation in a fluid of a finite depth is derived. The wave exciting force of the cylinder to an oscillation is also calculated by using the above relation. The method is applied to a circular cylinder swaying in a water of finite depth, and then, to a rectangular cylinder heaving, swaying, and rolling. The results of above cases give a good agreement with those by earlier investigators such as Bai, Keil, and Yeung. Also, this method is applied to a Lewis form cylinder with a half beam-to-draft ratio of 1.0 and a sectional area coefficient of 0.941, and to a bulbous section cylinder which is hard to represent by a mapping function. The results reveal that the hydrodynamic forces in heave increase as the depth of a water decrease, but in sway or roll, the tendency of the hydrodynamic forces is difficult to say in a few words. The exciting force to heave for a bulbous section cylinder becomes zero at two frequencies. The added mass moment of inertia for roll is seemed to mainly depend on the sectional shape than the water depth.

• Journal of the Korean Data and Information Science Society
• /
• 제9권2호
• /
• pp.345-355
• /
• 1998

다층전방향 신경망을 학습시키기 위해 역전파 알고리즘이 널리 사용되고 있으나 이 알고리즘은 긴 훈련시간, 극소점 문제, 이상치에 민감하다는 단점을 가지고 있다. 한편 실제문제에서는 많은 경우에 자료에 과대오차와 이상치가 포함되게 된다. 따라서 과대 오차에 민감하지 않고, 이상치의 영향을 최소화시키는 로버스트 역전파 알고리즘의 필요성이 대두되었다. 본 논문에서는 기존의 두종류의 로버스트 역전파 알고리즘을 이론적으로 비교하고 비선형 회귀 함수추정과 문자인식과 같은 패턴인식 문제에 적용하여 실험결과를 분석한다. 그리고 향후 연구과제로 신경망 학습을 위해 베이지안 기법의 사용을 제안한다.

• 한국측량학회지
• /
• 제35권5호
• /
• pp.423-430
• /
• 2017

Recently, data analysis research has been carried out using the deep learning technique in various fields such as image interpretation and/or classification. Various types of algorithms are being developed for many applications. In this paper, we propose a precipitation prediction algorithm based on deep learning with high accuracy in order to take care of the possible severe damage caused by climate change. Since the geographical and seasonal characteristics of Korea are clearly distinct, the meteorological factors have repetitive patterns in a time series. Since the LSTM (Long Short-Term Memory) is a powerful algorithm for consecutive data, it was used to predict precipitation in this study. For the numerical test, we calculated the PWV (Precipitable Water Vapor) based on the tropospheric delay of the GNSS (Global Navigation Satellite System) signals, and then applied the deep learning technique to the precipitation prediction. The GNSS data was processed by scientific software with the troposphere model of Saastamoinen and the Niell mapping function. The RMSE (Root Mean Squared Error) of the precipitation prediction based on LSTM performs better than that of ANN (Artificial Neural Network). By adding GNSS-based PWV as a feature, the over-fitting that is a latent problem of deep learning was prevented considerably as discussed in this study.

• 한국정밀공학회지
• /
• 제25권10호
• /
• pp.84-91
• /
• 2008

Due to the complexity of the engineering problems, it is difficult to obtain directly the stress field around the crack tip by theoretical derivation. In the paper, the hybrid method is employed to calculate full-field stress around the crack tip in uni-axially leaded finite width tensile plate, using the displacement data of given points calculated by finite element method as input data. The method uses complex variable formulations involving conformal mappings and analytical continuity. In order to accurately compare calculated fringes with experimental ones, both actual and reconstructed photoelastic fringe patterns are two times multiplied and sharpened by digital image processing. Reconstructed fringes by hybrid method are quite comparable to actual fringes. The experimental results indicate that Mode I stress intensity factor analyzed by the hybrid method are accurate within a few percent compared with ones obtained by empirical equation and finite element analysis.