• Journal of the Korean Data and Information Science Society
• /
• 제12권2호
• /
• pp.47-55
• /
• 2001

직접(direct) 추정법, 합성(synthetic) 추정법, 복합(composite) 추정법, 베이즈(Bayes) 추정법 등 소지역 추정법들의 효율성을 비교, 분석하고자 '98 경제활동 인구조사에서 경기도의 실제 자료를 이용하여 각 시부지역의 실업자수 추정값의 편의(bias)와 평균제곱 오차(MSE)를 모의실험을 통해 계산하였다.

• 한국통신학회논문지
• /
• 제39A권8호
• /
• pp.467-473
• /
• 2014

직접 수열 확산 대역 시스템에서는 (direct sequence spread spectrum: DSSS) 올바른 신호 동기화가 매우 중요하며, 이에 따라 부호 획득을 위한 다양한 순차 추정 기반 기법들이 연구되어 왔다. 대표적으로, rapid acquisition sequential estimation (RASE), seed accumulating SE (SASE), recursive soft SE (RSSE) 등의 기법이 연구되었다. 하지만, 기존의 기법들 간의 객관적인 성능 비교 및 분석은 현재까지 이루어진 바 없다. 본 논문에서는 순차 추정 기반 부호 획득 기법의 대표적 성능 지표인 올바른 칩 추정 확률 및 평균 부호 획득 시간을 (MAT) 이용하여 RASE, SASE, 및 RSSE 기법의 성능을 비교 및 분석한다.

• 한국물환경학회지
• /
• 제23권4호
• /
• pp.474-481
• /
• 2007

With population growth, industrialization, and urbanization within the watershed, the hydrologic response changed dramatically, resulting in increases in peak flow with lesser time to peak and total runoff with shortened time of concentration. Infiltration is directly affected by initial soil moisture condition, which is a key element to determine runoff. Influence of the initial soil moisture condition on hydrograph analysis should be evaluated to assess land use change impacts on runoff and non-point source pollution characteristics. The Long-Term Hydrologic Impact Assessment (L-THIA) model has been widely used for the estimation of the direct runoff worldwide. The L-THIA model was applied to the Little Eagle Creek (LEC) watershed and Its estimated direct runoff values were compared with the BFLOW filtered direct runoff values by other researchers. The $R^2$ value Was 0.68 and the Nash-Sutcliffe coefficient value was 0.64. Also, the L-THIA estimates were compared with those separated using optimized $BFI_{max}$ value for the Eckhardt filter. The $R^2$ value and the Nash-Sutcliffe coefficient value were 0.66 and 0.63, respectively. Although these higher statistics could indicate that the L-THIA model is good in estimating the direct runoff reasonably well, the Antecedent Moisture Condition (AMC) was not adjusted in that study, which might be responsible for mismatches in peak flow between the L-THIA estimated and the measured peak values. In this study, the L-THIA model was run with AMC adjustment for direct runoff estimation. The $R^2$ value was 0.80 and the Nash-Sutcliffe coefficient value was 0.78 for the comparison of L-THIA simulated direct runoff with the filtered direct runoff. However there was 42.44% differences in the L-THIA estimated direct runoff and filtered direct runoff. This can be explained in that about 80% of the simulation period is classified as 'AMC I' condition, which caused lower CN values and lower direct runoff estimation. Thus, the coefficients of the equation to adjust CN II to CN I and CN III depending on AMC condition were modified to minimize adjustments impacts on runoff estimation. The $R^2$ and the Nash-Sutcliffe coefficient values increase, 0.80 and 0.80 respectively. The difference in the estimated and filtered direct runoff decreased from 42.44% to 7.99%. The results obtained in this study indicate the AMC needs to be considered for accurate direct runoff estimation using the L-THIA model. Also, more researches are needed for realistic adjustment of the AMC in the L-THIA model.

• 한국수자원학회논문집
• /
• 제42권11호
• /
• pp.1017-1028
• /
• 2009

SRPM (Neyman-Scott Rectangular Pulse Model)은 수문학분야에서 널리 쓰이고 있는 강수생성모형이다. NSRPM을 구축하기 위해서는 총 5개의 매개변수를 추정하여야 한다. 일반적으로 사용되는 모멘트를 이용하여 매개변수를 추정할 경우, 사용되는 목적함수의 증가에 따라 추정되는 매개변수의 결과가 평탄해지고 목적함수를 추가하거나 조정하기 위해서는 복잡한 수식을 다시 계산해야 하며 추정된 매개변수가 무작위변수 생성 모형에 따라 상이한 결과를 나타내는 단점이 있다. 본 연구에서는 직접적인 매개변수 추정방법을 제시하여 모멘트를 이용한 매개변수 추정의 단점을 극복하고자 하였다. 직접적인 추정방법을 적용하기 위하여 NSRPM의 강수 생성 개수에 따른 통계치 변화를 모의하여 직접적인 추정을 위한 모형을 구축하였다. 기상청 청주 지상관측소의 관측 강수 자료를 사용하여 모멘트를 이용하여 추정된 매개변수와 직접적인 방법을 이용하여 매개변수를 추정하였다. 총 4 개의 무작위변수 알고리즘을 적용하여 강수를 생성하였고, 관측 강수 시계열을 이용하여 정확도를 비교하였다. 비교 결과 직접적인 추정방법이 모멘트를 이용한 매개변수 추정방법보다 안정적이고 높은 정확도를 보이는 매개변수를 추정하는 것으로 나타났다.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2002년도 Proceedings of International Symposium on Remote Sensing
• /
• pp.507-515
• /
• 2002

Remote Sensing Monitoring and Loss Estimated System of Flood Disaster based on GIS is an integrated system comprised flood disaster information receiving and collection, flood disaster simulation, and flood disaster estimation. When the system receives and collects remote sensing monitoring and conventional investigation information, the distributional features of flood disaster on space and time is obtained by means of image processing and information fusion. The economic loss of flood disaster can be classified into two pus: direct economic loss and indirect economic loss. The estimation of direct economic loss applies macroscopic economic analysis methods, i.e. applying Product (Industry and Agriculture Gross Product or Gross Domestic Product - GDP) or Unit Synthetic Economic Loss Index, direct economic loss can be estimated. Estimating indirect economic loss applies reduction coefficient methods with direct economic loss. The system can real-timely ascertains flood disaster and estimates flood Loss, so that the science basis fur decision-making of flood control and relieving disaster may be provided.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1992년도 하계학술대회 논문집 A
• /
• pp.377-380
• /
• 1992

This paper proposes a novel direct adaptive pole placement control algorithm which can be applied to continuous time nonminimum phase systems. The algorithm is based on Lyapunov's direct method. By introducing an auxiliary signal, a minimal error model is constructed in state space. Using the error model an estimation law is obtained via Lyapunov's second stability theorem. The global stability of the overall system is established.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1989년도 하계종합학술대회 논문집
• /
• pp.70-74
• /
• 1989

To track the target trajectory with maneuvers, unknown maneuvering inputs must be estimated. To do this the direct estimation algorithm using generalized least square technique is developed based on the procedure of failure detection and identification(FDI) theory. Through the simulation using maneuvering target scenario, tracking performance and efficiency of the algorithm developed here are investigated.

• Journal of the Korean Statistical Society
• /
• 제25권3호
• /
• pp.359-368
• /
• 1996

The maximum likelihood estimation is discussed for the NLAR model with Laplacian marginals. Since the explicit form of the estimates cannot be obtained due to the complicated nature of the likelihood function we utilize the automatic computer optimization subroutine using a direct search complex algorithm. The conditional least square estimates are used as initial estimates in maximum likelihood procedures. The results of a simulation study for the maximum likelihood estimates of the NLAR(1) and the NLAR(2) models are presented.

• 전력전자학회논문지
• /
• 제9권5호
• /
• pp.413-419
• /
• 2004

고피나스 모델 자속관측기를 사용한 유도전동기 직접벡터제어 시스템에서 회전자 시정수에 오차가 있을 경우 시스템의 동특성이 저하되는 현상을 살펴보았으며, 이를 해결하기 위한 방안으로 MRAC를 이용한 회전자 시정수 실시간 추정 알고리즘을 제안하였다. 제안한 방식의 타당성을 검증하기 위해서 자속관측기에 사용되는 전동기 상수들의 오차가 회전자 시정수 추정에 미치는 영향을 실험적으로 분석한 결과 전동기 상수에 다소 오차가 있더라도 회전자 시정수의 추정오차는 5% 이내로 수렴됨을 보였다. 따라서 본 연구에서 제안한 회전자 시정수 추정 알고리즘을 고피나스 모델 자속관측기에 적용하게 되면 전동기 상수 변동에 강인한 유도전동기 직접벡터제어 시스템을 구현할 수 있다.

• 대한임베디드공학회논문지
• /
• 제17권4호
• /
• pp.209-215
• /
• 2022

This paper proposes a novel approach for training an unsupervised depth estimation algorithm. The objective of unsupervised depth estimation is to estimate pixel-wise distances from camera without external supervision. While most previous works focus on model architectures, loss functions, and masking methods for considering dynamic objects, this paper focuses on the training framework to effectively use depth cue. The main loss function of unsupervised depth estimation algorithms is known as the photometric error. In this paper, we claim that direct depth cue is more effective than the photometric error. To obtain the direct depth cue, we adopt the technique of knowledge distillation which is a teacher-student learning framework. We train a teacher network based on a previous unsupervised method, and its depth predictions are utilized as pseudo labels. The pseudo labels are employed to train a student network. In experiments, our proposed algorithm shows a comparable performance with the state-of-the-art algorithm, and we demonstrate that our teacher-student framework is effective in the problem of unsupervised depth estimation.