• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 The Third Asian Fuzzy Systems Symposium
• /
• pp.118-121
• /
• 1998

In this paper, we propose another paradigm(QNBP) to be capable of overcoming Limitations of the traditional backpropagation(SDBP). QNBPis based on the method of Quasi -Newton(variable metric) with the nomalized direction vectors and computes step size through the linear search. Simulation results showed that QNBP was definitely superior to both the stochasitc SDBP and the deterministic SDBP in terms of accuracy and rate of convergence and might sumount the problem of local minima. and there was no different between DFP+SR1 and BFGS+SR1 combined algrothms in QNBP.

• 대한기계학회논문집
• /
• 제19권2호
• /
• pp.433-442
• /
• 1995

For the line search of a multi-variable optimization, an efficient algorithm is presented. The algorithm sequentially employs several polynomial approximations such as 2-point quadratic interpolation, 3-point cubic interpolation/extrapolation and 4-point cubic interpolation/extrapolation. The order of polynomial function is automatically increased for improving the accuracy of approximation. The method of approximation (interpolation or extrapolation) is automatically switched by checking the slope information of the sample points. Also, for selecting the initial step length along the descent vector, a new approach is presented. The performance of the proposed method is examined by solving typical test problems such as mathematical problems, mechanical design problems and dynamic response problems.

• 대한전자공학회논문지
• /
• 제26권9호
• /
• pp.1420-1427
• /
• 1989

A new block matching algorithm that improved the existing block matching algorithm in terms of image quality is proposed in this paper. The subblock of image including the vertical edge of object is subdivided into new two subblocks, and the moving vector found. The result of computer simulation shows on real image that the image quality by the algorithm becomes higher than that of the three step search algorithm by 1.1dB.

• Journal of Electrical Engineering and Technology
• /
• 제4권4호
• /
• pp.515-520
• /
• 2009

According to the operating characteristics of pump applications, they should exhibit high efficiency and energy saving capabilities throughout the whole operating process. A novel efficiency optimization control strategy is presented here to meet the high efficiency demand of a variable speed Permanent Magnet Synchronous Motor (PMSM). The core of this strategy is the excellent integration of mended maximum torque to the current control algorithm, based on the losses model during the dynamic and the grade search method with changed step by fuzzy logic during the steady. The performance experiments for the control system of a variable speed high efficiency PMSM have been completed. The test results verified that the system can reliably operate with a different control strategy during dynamic and steady operation, and the system exhibits better performance when using the efficiency-optimization control.

• 한국정보처리학회논문지
• /
• 제6권8호
• /
• pp.2271-2280
• /
• 1999

본고에서는 움직임 추정 성능을 개선하고 과도한 연산량과 전송 부담을 경감시키기 위해 HBMA에 기반한 가변 움직임 추정 기법을 제안한다. 제안된 알고리즘은 크게 다음과 같이 네 단계로 구성된다. 우선, 연속된 두 프레임 간의 차영상 윤곡 정보에서 정의한 블록 활동도를 평균하여 현재 영상의 평균 블록 활동도를 산출한다. 두 번째로, 이렇게 산출한 평균 블록 활동도를 통해 카메라 패닝의 유무를 검출한 후, 웨이블렛 변환에 의해 구성한 피라미드 계층 구조상에서 카메라 패닝 벡터를 추정하여 보상한다. 다음으로, 카메라 패닝 보상 후에 정의한 블록 활동도를 토대로 각 블록을 움직임 블록, 준 움직임 블록, 비 움직임 블록 중 어느 하나로 분류한 검색 테이블을 작성한다. 마지막으로, 제안된 가변 HBMA는 검색 테이블을 참조하여 블록 크기를 가변시키고 초기 탐색 계층 및 탐색 영역을 적응적으로 선정함으로써 피라미드 계층 구조상에서 효율적인 고속 움직임 추정을 수행할 수 있다. 이상에서 설명한 각 단계에서 요구되는 비용함수는 차영상 윤곽정보를 통해 획득한 블록 활동도를 공통적으로 이용한다.

• Nonlinear Functional Analysis and Applications
• /
• 제29권2호
• /
• pp.307-332
• /
• 2024

Two inertial extragradient-type algorithms are introduced for solving convex pseudomonotone variational inequalities with fixed point problems, where the associated mapping for the fixed point is a 𝜌-demicontractive mapping. The algorithm employs variable step sizes that are updated at each iteration, based on certain previous iterates. One notable advantage of these algorithms is their ability to operate without prior knowledge of Lipschitz-type constants and without necessitating any line search procedures. The iterative sequence constructed demonstrates strong convergence to the common solution of the variational inequality and fixed point problem under standard assumptions. In-depth numerical applications are conducted to illustrate theoretical findings and to compare the proposed algorithms with existing approaches.

• 품질경영학회지
• /
• 제50권1호
• /
• pp.77-89
• /
• 2022

Purpose: In order to improve the audit quality of a company, an in-depth analysis is required to categorize the audit report in the form of a text document containing the details of the external audit. This study introduces a systematic methodology to extract keywords for each group that determines the differences between groups such as 'audit plan' and 'interim audit' using audit reports collected in the form of text documents. Methods: The first step of the proposed methodology is to preprocess the document through text mining. In the second step, the documents are classified into groups using machine learning techniques and based on this, important vocabularies that have a dominant influence on the performance of classification are extracted. In the third step, the association rules for each group's documents are found. In the last step, the final keywords for each group representing the characteristics of each group are extracted by comparing the important vocabulary for classification with the important vocabulary representing the association rules of each group. Results: This study quantitatively calculates the importance value of the vocabulary used in the audit report based on machine learning rather than the qualitative research method such as the existing literature search, expert evaluation, and Delphi technique. From the case study of this study, it was found that the extracted keywords describe the characteristics of each group well. Conclusion: This study is meaningful in that it has laid the foundation for quantitatively conducting follow-up studies related to key vocabulary in each stage of auditing.

• 한국지능시스템학회논문지
• /
• 제27권1호
• /
• pp.22-28
• /
• 2017

Convolutional Neural Network(CNN)는 특징 추출과 분류의 두 단계로 나눌 수 있다. 그 중 특징 추출 단계의 커널의 크기, 채널의 수, stride 등의 hyperparameter는 CNN의 구조를 결정할 뿐만 아니라 특징을 추출하는 데에도 영향을 주기 때문에 CNN의 전체적인 성능에도 영향을 준다. 본 논문에서는 Parameter-Setting-Free Harmony Search(PSF-HS) 알고리즘을 이용하여 CNN의 특징 추출 단계에서의 hyperparameter를 최적화 하는 방법을 제안하였다. CNN의 전체 구조를 설정한 뒤 hyperparameter를 변수로 설정하였고 PSF-HS 알고리즘을 적용하여 hyperparameter를 최적화 하였다. 시뮬레이션은 MATLAB을 이용하여 진행하였고 CNN은 mnist 데이터를 이용하여 학습과 테스트를 했다. 총 500번 동안 변수를 업데이트했고 제안하는 방법을 이용하여 구한 CNN 구조 중 가장 높은 정확도를 가지는 구조는 99.28%의 정확도로 mnist 데이터를 분류하는 것을 확인할 수 있었다.

• 세라미스트
• /
• 제21권4호
• /
• pp.427-432
• /
• 2018

In this study, the aspherical lens for optical communications produced not with an one-step pneumatic type of external pressurization system (existed GMP process) but a constant weight of self-loaded mold put up to upper core. So the lens is molding with self-loaded weight molding and it calls Weight Molding process. In self-loaded molding process, we measured changes of center thickness molding lenses with each variable molding temperatures and time to find the effect of center of lens thickness to search key factors. As experimental results, the center thickness reach to targeted lenses step time value was changed drastically and it depends by molding temperature. If the molding temperature gets higher, the targeted lens that is reaching to the center thickness step time value was decreased. To find the effect of life improvement on mold core by imposing the self-loaded molding process we molded with GMP(Glass molding press) method and self-loaded molding method for 9,000 times and measured the lenses shape accuracy and surface roughness to evaluate the core life. As a result the self-loaded molding method core has 2,000 times longer that GMP (Glass molding press) method. If we adopt self-loaded molding method of the optical aspherical lens molding in the future, we expect that it would reduce the expense of changing the molds by molding core life improvements.

• 천문학회보
• /
• 제38권2호
• /
• pp.80-80
• /
• 2013

대규모 광도곡선 자료에서 다양한 주기변광성들의 정확한 주기를 효율적으로 검출하는 실험을 시도하였다. 실험을 위해 OGLE-III 맥동 변광성(RR Lyrae, Delta Scuti, Cepheid) 목록 중, I 필터로 관측된 총 31,324개의 광도 곡선을 사용하였다. 이 실험에 사용한 주기분석 알고리즘 MS_Period(Multi-Step period searching method)는 주기를 놓치지 않기 위해 두 가지 다른 방법(Multi Polynomial function, Phase Dispersion)으로 후보 주기를 구하고 정밀주기를 도출하기 위해 후보 주기 주변부를 Spline fitting을 통해 재탐색하는 방법이다. 기존의 MS_Period 방식은 주기 탐색 간격(dP/P)이 일정하였으나, 우리는 탐색 주기 구간을 나누고 짧은 주기에서는 작은 간격으로, 긴 주기에서는 보다 넓은 간격으로 주기를 탐색하는 과정을 추가하였다. 그 결과 98% 이상의 별에서 OGLE-III와 거의 일치하는 주기를 얻었으며, 긴 주기에서의 불필요한 정밀 탐색을 회피함으로써 분석시간도 단축되었다. 주기 결정이 어려운 경우들은 주로 1) periodogram에서 실제 주기가 아닌 1일 근처에서 noise보다 큰 peak가 보이는 경우, 2) 하나의 별에 대해 여러 주기가 비슷한 Phase diagram을 보이고, periodogram에서도 비슷한 peak를 갖는 경우, 3) OGLE-III의 주기와 전혀 다른 주기만 찾은 경우, 4) OGLE-III에서 제시하지 않은 혼합된 주기의 존재가 의심되는 경우인 것을 확인하였고, 각 사례들의 특징을 살펴보았다.