• Communications for Statistical Applications and Methods
• /
• 제9권3호
• /
• pp.809-824
• /
• 2002

we propose two methods which separate the variable selection step and the split-point selection step. We call these two algorithms as CHITES method and F&CHITES method. They adapted some of the best characteristics of CART, CHAID, and QUEST. In the first step the variable, which is most significant to predict the target class values, is selected. In the second step, the exhaustive search method is applied to find the splitting point based on the selected variable in the first step. We compared the proposed methods, CART, and QUEST in terms of variable selection bias and power, error rates, and training times. The proposed methods are not only unbiased in the null case, but also powerful for selecting correct variables in non-null cases.

• 방송공학회논문지
• /
• 제13권1호
• /
• pp.128-136
• /
• 2008

본 논문에서는 인접 블록 움직임 벡터의 통계적 특성을 이용한 가변 탐색 스텝의 고속 움직임 추정 기법에 대해 제안한다. 인접 블록 움직임 벡터들 사이의 상관 관계를 이용하여 움직임 추정을 위한 탐색 영역을 적응적으로 결정하였으며, 이를 통해 불필요한 탐색 지점 수를 제거할 수 있었다. 이와 같이 결정된 탐색 영역을 기반으로 가변 탐색 스텝 움직임 추정을 적용하였으며 움직임 추정을 위한 연산량을 줄일 수 있었다. 실험 결과를 통해 제안 방식이 H.264 JM의 고속 전 대역 spiral 탐색 기법과 기타 고속 움직임 추정 방식과 비교하여 부호화 성능의 저하 없이 움직임 추정을 위한 탐색 지점 수 및 연산 량이 급격히 감소됨을 확인할 수 있었다.

• 한국통신학회논문지
• /
• 제30권11C호
• /
• pp.1091-1097
• /
• 2005

본 논문에서는 H.264 동영상 표준 부호화 방식을 위한 고속 움직임 추정 기법을 제안한다. 제안된 고속 움직임 추정 기법은 움직임 벡터가 국부적으로 일정한 상관관계를 유지하고 있다는 특성을 이용하여 각 블록의 움직임 추정 시 부호화된 인접 블록의 움직임 벡터 값으로부터 주어진 탐색 영역 내에서 수평, 수직 방향 각각 독립적으로 불필요한 탐색 영역을 제거하여 축소된 비대칭 가변 크기 탐색 영역을 결정한다. 또한, 예측된 탐색 영역제약 조건을 이용하여 가변 단계 탐색 방식을 취하는 것을 특징으로 한다. 실험 결과를 통해 제안된 방식에 의해 움직임 추정 탐색 지점이 전영역 탐색 방식(Full Search)에 비해 평균 $98\%$ 이상 절감되었으며, PSNR과 비트율은 평균적으로 전영역 탐색 방식과 등일 수준인 것을 결과로부터 확인 할 수 있었다.

• Journal of Electrical Engineering and Technology
• /
• 제10권2호
• /
• pp.529-538
• /
• 2015

Due to the high efficiency and compact mechanical structure, direct drive variable speed generators are used for power conversion in wind turbines. The wind energy conversion system (WECS) considered in this paper consists of a permanent magnet synchronous generator (PMSG), uncontrolled rectifier, dc-dc boost converter controlled with maximum power point tracking (MPPT) and adaptive hysteresis controlled voltage source inverter (VSI). For high utilization of the converter's power capability and stabilizing voltage and power flow, constant DC-link voltage is essential. Step and search MPPT algorithm which senses the rectified voltage ($V_{DC}$) alone and controls the same is used to effectively maximize the output power. The adaptive hysteresis band current control is characterized by fast dynamic response and constant switching frequency. With MPPT and adaptive hysteresis band current control in VSI, the DC link voltage is maintained constant under variable wind speeds and transient grid currents respectively.

• ETRI Journal
• /
• 제45권1호
• /
• pp.119-130
• /
• 2023

We propose a quantized gradient search algorithm that can achieve global optimization by monotonically reducing the quantization step with respect to time when quantization is composed of integer or fixed-point fractional values applied to an optimization algorithm. According to the white noise hypothesis states, a quantization step is sufficiently small and the quantization is well defined, the round-off error caused by quantization can be regarded as a random variable with identically independent distribution. Thus, we rewrite the searching equation based on a gradient descent as a stochastic differential equation and obtain the monotonically decreasing rate of the quantization step, enabling the global optimization by stochastic analysis for deriving an objective function. Consequently, when the search equation is quantized by a monotonically decreasing quantization step, which suitably reduces the round-off error, we can derive the searching algorithm evolving from an optimization algorithm. Numerical simulations indicate that due to the property of quantization-based global optimization, the proposed algorithm shows better optimization performance on a search space to each iteration than the conventional algorithm with a higher success rate and fewer iterations.

• 한국정보시스템학회지:정보시스템연구
• /
• 제26권3호
• /
• pp.171-185
• /
• 2017

Purpose Recently, there has been an increase in attempts to analyze social phenomena, consumption trends, and consumption behavior through a vast amount of customer data such as web search traffic information and social buzz information in various fields such as flu prediction and real estate price prediction. Internet portal service providers such as google and naver are disclosing web search traffic information of online users as services such as google trends and naver trends. Academic and industry are paying attention to research on information search behavior and utilization of online users based on the web search traffic information. Although there are many studies predicting social phenomena, consumption trends, political polls, etc. based on web search traffic information, it is hard to find the research to explain and predict tourism demand and establish tourism policy using it. In this study, we try to use web search traffic information to explain the tourism demand for major cities in Gangwon-do, the representative tourist area in Korea, and to develop a nowcasting model for the demand. Design/methodology/approach In the first step, the literature review on travel demand and web search traffic was conducted in parallel in two directions. In the second stage, we conducted a qualitative research to confirm the information retrieval behavior of the traveler. In the next step, we extracted the representative tourist cities of Gangwon-do and confirmed which keywords were used for the search. In the fourth step, we collected tourist demand data to be used as a dependent variable and collected web search traffic information of each keyword to be used as an independent variable. In the fifth step, we set up a time series benchmark model, and added the web search traffic information to this model to confirm whether the prediction model improved. In the last stage, we analyze the prediction models that are finally selected as optimal and confirm whether the influence of the keywords on the prediction of travel demand. Findings This study has developed a tourism demand forecasting model of Gangwon-do, a representative tourist destination in Korea, by expanding and applying web search traffic information to tourism demand forecasting. We compared the existing time series model with the benchmarking model and confirmed the superiority of the proposed model. In addition, this study also confirms that web search traffic information has a positive correlation with travel demand and precedes it by one or two months, thereby asserting its suitability as a prediction model. Furthermore, by deriving search keywords that have a significant effect on tourism demand forecast for each city, representative characteristics of each region can be selected.

• 전자공학회논문지S
• /
• 제34S권9호
• /
• pp.77-83
• /
• 1997

We present a fast adaptive block matching algorithm using variable search area and subsampling to estimate motion vector more exactly. In the presented method, the block is classified into one of three motion categories: zero motion vector block, medium-motion bolck or high-motion block according to mean absolute difference of the block. By the simulation, the computation amount of the presented methoe comparable to three step search algorithm and new three step search algorithm. In the fast image sequence, the PSNR of our algorithm increased more than TSS and NTSS, because our algorithm estimated motion vector more accurately.

• Journal of applied mathematics & informatics
• /
• 제8권1호
• /
• pp.181-197
• /
• 2001

A heuristic method TABU-CSP using Tabu Search (TS) is described for solving Constraint Satisfaction Problems (CSPs). The method is started with a complete but inconsistent solution of a binary CSP and obtained in prespecified number of iterations either a consistent solution or a near optimal solution with an acceptable number of conflicts. The repair in the solution at each iterative step is done by using two heuristics alternatively. The first heuristic is a min-conflicts heuristic that chooses a variable with the maximum number of conflicts and reassigns it the value which leads to the minimum number of conflicts. If the acceptable solution is not reached after the search continued for a certain number of iterations, the min-conflict heuristic is changed and the variable selected least number of times is chosen for repair. If an acceptable solution is not reached, the method switches back to the min-conflict heuristic and proceeds further. This allowed the method to explore a different region of search space space for the solution as well as to prevent cycling. The demonstration of the method is shown on a toy problem [9]which has no solution. The method is then tested on various randomly generated CSPs with different starting solutions. The performance of the proposed method in terms of the average number of consistency is checked and the average number of conflicts is conflicts is compared with that of the Branch and Bound(BB) method used to obtain the same solution. In almost all cases, the proposed method moves faster to the acceptable solution than BB.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2008년도 하계종합학술대회
• /
• pp.285-286
• /
• 2008

In this paper, we propose adaptive search channel estimate algorithm. The proposed algorithm is modified LMS algorithm which has a variable step size and parallel convolution. In simulation result, a error estimate accuracy of the proposed algorithm is about -20 dB and general LMS algorithm is about 10 dB. The proposed algorithm is better error estimate accuracy than general LMS algorithm.

• 천문학회지
• /
• 제37권2호
• /
• pp.79-85
• /
• 2004

We developed an algorithm to identify and determine periods of variable sources. With its robustness and high speed, it is expected to become an useful tool for surveys with large volume of data. This new scheme consists of an initial coarse. process of finding several candidate periods followed by a secondary process of much finer period search. With this multi-step approach, best candidates among statistically possible periods are produced without human supervision and also without any prior assumption on the nature of the variable star in question. We tested our algorithm with 381 stars taken from the ASAS survey and the result is encouraging. In about $76\%$ cases, our results are nearly identical as their published periods. Our algorithm failed to provide convincing periods for only about $10\%$ cases. For the remaining $14\%$, our results significantly differ from their periods. We show that, in many of these cases, our periods are superior and much closer to the true periods. However, the existence of failures, and also periods sometimes worse than manually controlled results, indicates that this algorithm needs further improvement. Nevertheless, the present experiment shows that this is a positive step toward a fully automated period analysis for future variability surveys.