• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2020년도 추계학술발표대회
• /
• pp.850-853
• /
• 2020

gradient decent 를 기반으로 한 Differentiable architecture search(DARTS)는 한 번의 Architecture Search 로 모든 후보 연산 중 가장 가중치가 높은 연산 하나를 선택한다. 이 때 비슷한 종류의 연산이 가중치를 나누어 갖는 "표의 분산"이 나타나, 성능이 더 좋은 연산이 선택되지 못하는 상황이 발생한다. 본 연구에서는 이러한 상황을 막기위해 Architecture Parameter 가중치의 gradient 를 기반으로 연산들을 클러스터링 하여 그룹화 한다. 그 후 그룹별로 가중치를 합산하여 높은 가중치를 갖는 그룹만을 사용하여 한 번 더 Architecture Search 를 진행한다. 각각의 Architecture Search 는 DARTS 의 절반 epoch 만큼 이루어지며, 총 epoch 이 같으나 두번째의 Architecture Search 는 선별된 연산 그룹을 사용하므로 DARTS 에 비해 더 적은 Search Cost 가 요구된다. "표의 분산"문제를 해결하고, 2 번으로 나뉜 Architecture Search 에 따라 CIFAR 10 데이터 셋에 대해 2.46%의 에러와 0.16 GPU-days 의 탐색시간을 얻을 수 있다.

• 천문학회보
• /
• 제44권2호
• /
• pp.77.2-77.2
• /
• 2019

Measurements of supermassive black hole (SMBH) masses are crucial in studying the co-evolution of SMBHs and their host galaxies. Although reverberation mapping is the most accurate method known to date, this requires spectroscopic monitoring over long periods. Thus, the current sample barely reaches three digits. The virial method, on the other hand, uses emission-line and continuum properties from a single spectrum to estimate the SMBH mass; hence the name single-epoch method. The Sloan Digital Sky Survey (SDSS) has observed spectra of almost all quasi-stellar objects (QSOs) discovered so far. Building on previous studies, using the single-epoch method, we estimate the SMBH masses of more than 500,000 QSOs from the SDSS DR14 Quasar Catalog. This increases the mass-estimated SMBH sample almost by a factor of two, and especially more for the low-mass regime, which was the main target of SDSS-IV (eBOSS).

• 천문학회보
• /
• 제44권2호
• /
• pp.78.3-78.3
• /
• 2019

We develop a novel technique that can constrain the evolutionary track of the epoch of reionization (EoR) by applying the convolutional neural network (CNN) to the 21-cm differential brightness temperature. We use 21cmFAST, a fast semi-numerical cosmological 21-cm signal simulator, to produce mock 21-cm map between z=6-13. We design a CNN architecture that predicts the volume-averaged neutral hydrogen fraction from the given 21-cm map. The estimated neutral fraction has a good agreement with its truth value even after smoothing the 21-cm map with somewhat realistic choices of beam size and the frequency bandwidth of the Square Kilometre Array (SKA). Our technique could be further utilized to denoise the 21-cm map or constrain the properties of the radiation sources.

• 천문학회보
• /
• 제45권1호
• /
• pp.66.2-66.2
• /
• 2020

We propose a deep learning analysis technique with a convolutional neural network (CNN) to predict the evolutionary track of the Epoch of Reionization (EoR) from the 21-cm differential brightness temperature tomography images. We use 21cmFAST, a fast semi-numerical cosmological 21-cm signal simulator, to produce mock 21-cm maps between z = 6 ~ 13. We then apply two observational effects, such as instrumental noise and limit of (spatial and depth) resolution somewhat suitable for realistic choices of the Square Kilometre Array (SKA), into the 21-cm maps. We design our deep learning model with CNN to predict the sliced-averaged neutral hydrogen fraction from the given 21-cm map. The estimated neutral fraction from our CNN model has great agreement with the true value even after coarsely smoothing with broad beam size and frequency bandwidth and heavily covered by noise with narrow beam size and frequency bandwidth. Our results show that the deep learning analyzing method has the potential to reconstruct the EoR history efficiently from the 21-cm tomography surveys in future.

• 한국의상디자인학회:학술대회논문집
• /
• 한국의상디자인학회 2005년도 Fasion road:Korea to belarus
• /
• pp.17-18
• /
• 2005

• 자원환경지질
• /
• 제19권2호
• /
• pp.153-163
• /
• 1986

Paleomagnetic and geological studies of volcanic and sedimentary rocks of Jeju Island have been carried out to determine the position of virtual geomagnetic pole(VGP), and to estimate the geological sequence and their age. As a result of paleomagnetic studies, the reversal polarities are measured in the Sanbangsan trachyte and Hwasun formation, and the normal are the rest. In case of normal polarity, the mean values of declination and inclination are $2.3^{\circ}$ and $48.4^{\circ}$, respectively, and the average value of VGP is $85.4^{\circ}N$ and $79.9^{\circ}W$. The locations of VGP's are coincident with those obtained from world-wide Plio-Pleistocene rocks. The Hwasun formation and Seongsan formation which have been known to be sedimented in the similar time in the 2nd-stage of volcanic eruption, possess reversal and normal polarities, respectively. This fact brings about the result that two formations should be separated in a sense of geological sequence. Consequently, the geological sequence of the 2nd-stage of volcanic eruption is Pyoseonri basalt-Seoguipo hawaiite-Hwasun formation-Seongsan formation-Jungmun hawaiite-Sanbangsan trachyte. Referring to the paleomagnetic studies and the previous and present geological studies, Seoguipo formation corresponds to the Gauss normal epoch, the 2nd-stage of volcanic eruption to Matuyama reversed epoch, and the 3rd-, 4th-, and 5th-stages to Brunhes normal epoch. Therefore, the Seoguipo formation is mostly sedimented during late Pliocene and/or presumably extended to the early Pleistocene. The rocks of the 2nd- to 5th-stage are formed later than this.

• 천문학논총
• /
• 제30권2호
• /
• pp.61-64
• /
• 2015

The brightness of Io's magnetic footprint, an indicator of electromagnetic interaction at the satellite, appears to be strongly connected to the satellite's distance from the plasma equator. As a result, the brightest footprints were detected when Io is near the interception location between the satellite's orbital plane and the plasma equator. However, volcanic activities on Io show strong correlation with the equatorward shift of Jupiter's main auroral oval, consequently causing the disappearance of Io's footprint. The same conclusion was suggested via the observation of Jupiter's hectometric radio emission, called HOM, which closely corresponds to Jupiter's auroral activity. The plasma environment near the Jovian satellites was found to vary significantly at different observational epochs. The electron density increased by approximately a factor of three from the Voyager epoch (1979) to the Galileo epoch (1995), while the electron density was found to be significantly higher (~ 5 times) in the Cassini epoch (2001). In this current study, the magnetic footprints were clearly brighter ten years ago (from peak brightness in 1998-2001) than the footprints detected in 2007. For volcanic activities on Io in 2007, there are two clear activities in February and late May. The magnetic footprint appeared to be dimmer in March 2007, expected to be the result of volcano activities in Feb 2007. However, the magnetic footprint brightness in June appeared to be slightly brighter than the footprints observed in May. The reason could be the time delay between the brightening of the sodium nebula on approximately May 31st and, a while later, the enhancement of flux tube content peaking on approximately June 5th. On the other hand, Io's magnetic footprints were observed during June 1st - 10th when they may not yet have been affected by the increase in mass outflow due to the increase of plasma density.

• 한국측량학회지
• /
• 제31권1호
• /
• pp.11-22
• /
• 2013

EAREF(East-Asia Reference Frame)는 지각판이 비교적 안정된 유라시아 플레이트를 기초로 동북아시아와 동남아시아를 포괄하는 동아시아 지역의 GNSS CORS 지구관측 네트워크를 구축하고 특정한 기준시점에서 좌표기준계를 관리하여 공간정보의 통합을 지원하는 것을 목표로 하고 있다. 본 연구에서는 동일본 대지진 이후인 2012년 1월 1일(2012.0) 기준시점에 대한 동아시아 지역의 EAREF의 정밀좌표 산정을 목적으로 하고 있으며, 데이터셋 구축에 대한 1차 연구에 이어 데이터 처리의 정확도 향상을 위한 다양한 실험을 거쳐 데이터 처리 방식을 도출하였다 그리고 이를 적용하여 동아시아 지역에 위치한 GNSS 상시관측망 EAREF의 2012.0 기준시점에 대한 정밀좌표의 초기결과를 산출하였다. 산출된 EAREF의 2012.0 시점의 정밀좌표는 IGS해석센터의 주간해와 X, Y, Z방향에서 평균적으로 각각 0.004 m, 0.007 m, 0.009 m의 차이가 나타났다. 또한, 향후 진행해야 할 EAREF 네트워크의 구현방향에 대하여 고찰하였다.

• 한국방사선학회논문지
• /
• 제14권1호
• /
• pp.39-44
• /
• 2020

본 연구에서는 CNN과 빅데이터 기술을 이용한 Deep Learning을 통해 흉부 X-ray 영상 분류 및 정확성 연구에 대하여 알아보고자 한다. 총 5,873장의 흉부 X-ray 영상에서 Normal 1,583장, Pneumonia 4,289장을 사용하였다. 데이터 분류는 train(88.8%), validation(0.2%), test(11%)로 분류하였다. Convolution Layer, Max pooling layer pool size 2×2, Flatten layer, Image Data Generator로 구성하였다. Convolution layer가 3일 때와 4일 때 각각 filter 수, filter size, drop out, epoch, batch size, 손실함수 값을 설정하였다. test 데이터로 Convolution layer가 4일 때, filter 수 64-128-128-128, filter size 3×3, drop out 0.25, epoch 5, batch size 15, 손실함수 RMSprop으로 설정 시 정확도가 94.67%였다. 본 연구를 통해 높은 정확성으로 분류가 가능하였으며, 흉부 X-ray 영상뿐만 아니라 다른 의료영상에서도 많은 도움이 될 것으로 사료된다.

• 대한한의학회지
• /
• 제16권2호
• /
• pp.44-61
• /
• 1995

Arranging 63 kinds of separate volumes and papers published on Oriental medicine, I could get the result as follows. 1. The herbs for internal application used commonly in vitiligo are 155 kinds totally. The herbs for external application are 67 kinds. The herbs for external and internal both application examined into 23 sorts. 2. Herbs for weakness syndrome in vitiligo are 49 kinds. Another type, excess syndrome is 105 sorts, the'latter is roughly twice as many as the former. 3. It is as follows that the results of study in relations to kinds, factors, and medical treatments of herbs about vitiligo in and out of the country with the division of former times to 1900, 1901 to 1980, 1981 to 1990, and 1991 to the latest day. In comparative study of inner and outer of thc country about factors and medical treatments of vitiligo in chronicle classification, its factors in the internal documents are classified by outside factors. Although there arc the differences of factors by each epoch, however, the factors of vitiligo according to external documents are blood stasis(血熱), deficiency of um of the liver and kidneys(肝腎陰虛), deficiency of blood(血虛), excess of exhaustion(勞倦過多) etc. Moreover, the medical treatment is more diverse and the differences by each epoch as to the medical treatment is also put down more saliantiy than in internal documents. 4. In comparison with herbs in experimental and no experimental documents, herbs applied for weakness syndromes in experimental method are 40 kinds totally. The herbs in no experimental methods are 35 kinds. The herbs used by experimental method are 65 kinds. The common herbs for excess syndrome by no experimental method are 78 kinds. We can see comparable difference from kinds of herbs used by experimental method. In brief, there are the differences classified by each epoch in Oriental medicine for treatment. Especially one of the most important feature, the frequency in use of weakness syndrome herbs has increased more than that of excess syndrome herbs. In external documents (china) and experimental study, generally the differences of common herbs and factors have disappeared through many experimentsitudy. The classification of its factors have been fractionalized clinically. Besides, in Western medicine and Orienal medicine, vitiligo tends to be prescribed not to simple skin disease but a mental and physical disease, a whole body and an internal disease.