http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Revisiting the Bradley-Terry model and its application to information retrieval
전종준,김용대 한국데이터정보과학회 2013 한국데이터정보과학회지 Vol.24 No.5
The Bradley-Terry model is widely used for analysis of pairwise preference data. We explain that the popularity of Bradley-Terry model is gained due to not only easy com-putation but also some nice asymptotic properties when the model is misspecified. For information retrieval required to analyze big ranking data, we propose to use a pseudo likelihood based on the Bradley-Terry model even when the true model is different from the Bradley-Terry model. We justify using the Bradley-Terry model by proving that the estimated ranking based on the proposed pseudo likelihood is consistent when the true model belongs to the class of Thurstone models, which is much bigger than the Bradley-Terry model.
문상준,전종준 한국데이터정보과학회 2017 한국데이터정보과학회지 Vol.28 No.6
The online learning is a process of obtaining the solution for a given objective function where the data is accumulated in real time or in batch units. The stochastic gradient descent method is one of the most widely used for the online learning. This method is not only easy to implement, but also has good properties of the solution under the assumption that the generating model of data is homogeneous. However, the stochastic gradient method could severely mislead the online-learning when the homogeneity is actually violated. We assume that there are two heterogeneous generating models in the observation, and propose the a new stochastic gradient method that mitigate the problem of the heterogeneous models. We introduce a robust mini-batch optimization method using statistical tests and investigate the convergence radius of the solution in the proposed method. Moreover, the theoretical results are confirmed by the numerical simulations. 온라인 학습은 자료가 실시간으로 혹은 배치 단위로 축적되는 상황에서 주어진 목적함수의 해를 계산하는 방법을 말한다. 온라인 학습 알고리즘 중 배치를 이용한 확률적 경사 하강법 (stochastic gradient decent method) 은 가장 많이 사용되는 방법 중 하나다. 이 방법은 구현이 쉬울 뿐만 아니라 자료가 동질적인 분포를 따른다는 가정 하에서 그 해의 성질이 잘 연구되어 있다. 하지만 자료에 특이값이 있거나 임의의 배치가 확률적으로 이질적 성질을 가질 때, 확률적 경사 하강법이 주는 해는 큰 편이를 가질 수 있다. 본 연구에서는 이러한 비정상 배치 (abnormal batch) 있는 자료 하에서 효과적으로 온라인 학습을 수행할 수 있는 수정된 경사 하강 알고리즘 (modified gradient decent algorithm)을 제안하고, 그 알고리즘을 통해 계산된 해의 수렴성을 밝혔다. 뿐만 아니라 간단한 모의 실험을 통해 제안한 방법의 이론적 성질을 실증하였다.
김광수,전종준,최호식 한국자료분석학회 2009 Journal of the Korean Data Analysis Society Vol.11 No.6
다른 변수에 의한 효과를 제거한, 두 변수간의 순수한 선형관계를 살펴볼 수 있는 편상관계수는 상관계수와 더불어 자료분석의 기본적인 방법으로써 널리 사용되고 있다. 그러나, 편상관계수는 정적인 상관관계를 가정을 바탕으로 하고 있기 때문에, 시간에 따라 변하는 동적인 관계를 파악하는 것에는 어느 정도의 제약성이 따른다. 본 논문에서는 두 변수의 상관성에 대해서 다른 변수가 미치는 동적인 영향을 정량적으로 분석할 수 있는 모형을 제안하고자 한다. 이변량 상관성 모형화에 대한 방법으로는 모수적 일반화선형모형과 B-스플라인 기저방법을 활용한 비모수적 Varying Coefficient 회귀모형을 고려하였다. 모의실험을 통해 제안한 두 방법을 평균제곱오차 등의 기준에서 비교한 결과, 모형이 비선형성 강할수록 Varying Coefficient에 근거한 방법이 우수함을 확인하였다. 또한, 실제 환율자료분석에 제안한 모형들을 적용한 결과, 미환율이 높을수록 엔화의 환율변동과 위안화의 환율변동간의 상관성이 증가하는 패턴을 파악할 수 있었다. The partial correlation coefficient is wildly used as a measure which quantifies pure linear correlation between two variables. Since the traditional partial correlation is based on the static distribution, however, it is difficult to model a dynamic partial correlation. In this paper, we consider a regression framework for modeling such a partial correlation. For this, we consider two models which are constructed by Generalized Linear Model and Varying Coefficient Model based on B-spline method. From two simulated data sets, as true model has more nonlinear structure, Varying Coefficient Model performs better than Generalized Linear Model in mean squared error rate. Also, the result of the analysis about foreign exchange rates shows that the correlation of Yen and Yuan exchange rates grows as Dollar exchange rate is higher.
도로교통소음과 도시구성요소 관계 분석을 위한 인공신경망 모형
김필립,류훈재,전종준,장서일 한국소음진동공학회 2019 한국소음진동공학회 논문집 Vol.29 No.5
Road-traffic noise is a critical factor that affects the life and health environments of urban inhabitants. In Korea, noise maps of cities created by commercial noise mapping software are used to manage road-traffic noise. This makes the management of noisy environments easy, but in the case of metropolitan cities, the creation of noise maps is time-consuming and costly. In this study, the relationship between road-traffic noise and urban form indicators (i.e., population, roads, buildings, and land use), showing the characteristics of a city, were analyzed to predict the road-traffic noise level using a statistical model. The road-traffic noise level predicted by the artificial neural network method was compared to that using the ordinary least squares method: The adjusted coefficient of determination (R2) of the former method was 0.5, while that of the latter model was 0.44. Furthermore, the floor space index was used as the urban form indicator, which has the largest effect on the road-traffic noise level.
성장현,박준형,전종준,서승범 대한토목학회 2020 KSCE JOURNAL OF CIVIL ENGINEERING Vol.24 No.9
Although many studies have sought to characterize future meteorological droughts, a few efforts have been done for quantifying the uncertainty, inter-model variability, arises from global circulation models (GCM) ensemble. A clear understanding of the uncertainty in multiple GCMs should be preceded before future meteorological droughts are projected. Therefore, this study evaluates the uncertainty in future meteorological drought characteristics that are induced by GCM ensemble using the custom measure “the degree of GCM spreading”. Future meteorological drought indices, the standardized precipitation index (SPI) and standardized precipitation evapotranspiration index (SPEI), were computed to five different time scales: 3, 6, 9, 12 and 24 months using statistically downscaled 28 GCMs under Representative Concentration Pathway (RCP) 4.5 and 8.5 at 60 weather stations in South Korea. The frequency, duration, and severity of drought events were estimated for three different future periods; F1 (2010 2039), F2 (2040 2069), and F3 (2070 2099). It was found that the uncertainty increases as the time scale lengthens regardless of a choice of drought indices or RCP scenarios. It also turned out that the SPI exhibits larger uncertainty rather than the SPEI, because temperature data exhibit a relatively much smaller variability comparing to precipitation data. Moreover, there was a shift of regions having larger values of the increasing rate between F1 and F2, which is shift from the north-western to southern region of South Korea.