• 센서학회지
• /
• 제25권6호
• /
• pp.431-434
• /
• 2016

Currently, the radial basis function network (RBFN) and various other neural networks are employed to classify gases using chemical sensors arrays, and their performance is steadily improving. In particular, the identification performance of the RBFN algorithm is being improved by optimizing parameters such as the center, width, and weight, and improved algorithms such as the radial basis function network-stochastic gradient (RBFN-SG) and radial basis function network-normalized stochastic gradient (RBFN-NSG) have been announced. In this study, we optimized the number of centers, which is one of the parameters of the RBFN-NSG algorithm, and observed the change in the identification performance. For the experiment, repeated measurement data of 8 samples were used, and the elbow method was applied to determine the optimal number of centers for each sample of input data. The experiment was carried out in two cases(the only one center per sample and the optimal number of centers obtained by elbow method), and the experimental results were compared using the mean square error (MSE). From the results of the experiments, we observed that the case having an optimal number of centers, obtained using the elbow method, showed a better identification performance than that without any optimization.

• 한국정보통신학회논문지
• /
• 제10권6호
• /
• pp.1032-1037
• /
• 2006

채널은 무선망에 있어서 한정된 주요 자원 중의 하나이다. 다양한 채널 관리 기법들이 제시되어 왔으며, 최근 들어 가드채널의 최적화 문제가 부각되고 있다. 본 논문에서는 신경망을 이용한 지능적인 채널 관리 기법을 제안한다. 신경망의 학습 데이터 생성과 성능분석을 위하여 SRN(Stochastic Reward Net) 채널 할당 모델이 개발된다. 제안된 기법에서 신경망은 지도학습 방법인 역전파 알고리즘을 이용하여 최적의 가드채널 값 g를 계산하도록 학습한다. 학습된 신경망을 이용하여 최적의 g를 계산하고, 이를 SRM모델에서 구해진 결과와 비교한다. 실험 결과는 신경망에서 구한 가드채널 수와 SRM모델로부터 구한 가드채널 수의 상대적 차이가 없음을 보여준다.

• International journal of advanced smart convergence
• /
• 제12권2호
• /
• pp.56-66
• /
• 2023

The goal of deep learning is to extract complex features from multidimensional data use the features to create models that connect input and output. Deep learning is a process of learning nonlinear features and functions from complex data, and the user data that is employed to train deep learning models has become the focus of privacy concerns. Companies that collect user's sensitive personal information, such as users' images and voices, own this data for indefinite period of times. Users cannot delete their personal information, and they cannot limit the purposes for which the data is used. The study has designed a deep learning method that employs privacy protection technology that uses distributed collaborative learning so that multiple participants can use neural network models collaboratively without sharing the input datasets. To prevent direct leaks of personal information, participants are not shown the training datasets during the model training process, unlike traditional deep learning so that the personal information in the data can be protected. The study used a method that can selectively share subsets via an optimization algorithm that is based on modified distributed stochastic gradient descent, and the result showed that it was possible to learn with improved learning accuracy while protecting personal information.

• 산업경영시스템학회지
• /
• 제39권4호
• /
• pp.137-146
• /
• 2016

In recent years, business environment is faced with multi uncertainty that have not been suffered in the past. As supply chain is getting expanded and longer, the flow of information, material and production is also being complicated. It is well known that development service industry using application software has various uncertainty in random events such as supply and demand fluctuation of developer's capcity, project effective date after winning a contract, manpower cost (or revenue), subcontract cost (or purchase), and overrun due to developer's skill-level. This study intends to social contribution through attempts to optimize enterprise's goal by supply chain management platform to balance demand and supply and stochastic programming which is basically applied in order to solve uncertainty considering economical and operational risk at solution supplier. In Particular, this study emphasizes to determine allocation of internal and external manpower of developers using S&OP (Sales & Operations Planning) as monthly resource input has constraint on resource's capability that shared in industry or task. This study is to verify how Stochastic Programming such as Markowitz's MV (Mean Variance) model or 2-Stage Recourse Model is flexible and efficient than Deterministic Programming in software enterprise field by experiment with process and data from service industry which is manufacturing software and performing projects. In addition, this study is also to analysis how profit and labor input plan according to scope of uncertainty is changed based on Pareto Optimal, then lastly it is to enumerate limitation of the study extracted drawback which can be happened in real business environment and to contribute direction in future research considering another applicable methodology.

• 한국지능시스템학회논문지
• /
• 제22권3호
• /
• pp.386-393
• /
• 2012

최근에 금융공학 분야에 보고된 바 있는 확률적 구간이동 기반 포트폴리오 선정기법은, 최적 포트폴리오 선정을 수행하는 과정에서 부(wealth)의 변화에 대한 동적 특성 및 여러 제약조건(constraints)을 명시적으로 고려할 수 있는 방법이다. 확률적 구간이동 최적화 기반 포트폴리오 선정기법은, 그동안 구간이동 최적화 기법이 다수의 공학 문제에서 성취하였던 이론적 가치, 범용성 및 효용 등을 고려할 때 현대 포트폴리오 이론 분야에서 또 하나의 주요한 기술혁신이 될 가능성을 가지고 있다. 이에 본 논문에서는 이론적 고찰을 바탕으로 단순화된 SDP 기반 동적 포트폴리오 선정이 가능함을 관찰하고, 이를 한국 주식시장에 적용하는 시뮬레이션 연구를 수행하여 결과 수익률에 관한 의미 있는 성과를 거두었다.

• 한국자동차공학회논문집
• /
• 제22권2호
• /
• pp.100-106
• /
• 2014

Lower control arm (LCA) is a part of chassis in automotive. Performances of LCA such as stiffness, durability and permanent displacement must be considered in design optimization. However it is hard to consider different performances at once in optimization because these are measured by different commercial tools like Radioss, Abaqus, etc. In this paper, firstly, we construct the integrated design automation system for LCA based on Matlab including Hypermesh, Radioss and Abaqus. Secondly, Akaike information criterion (AIC) is used for assessment of reliability of LCA. It can find the best estimated distribution of performance from limited and discrete stochastic information and then obtains the reliability from the distribution. Finally, we consider tolerances of design variables and variation of elastic modulus and achieve the target reliability by carrying out reliability-based design optimization (RBDO) with the integrated system.

• 한국지진공학회논문집
• /
• 제11권6호
• /
• pp.1-14
• /
• 2007

인접 구조물의 지진응답 제어를 위한 비선형 감쇠시스템의 최적 설계 방법에 관하여 연구하였다. 최적 설계를 위한 목적 함수로는 구조물의 응답과 감쇠기의 총 사용량을 고려하였으며, 상충하는 두 목적함수를 합리적인 수준에서 동시에 최소화하는 해를 구하기 위하여 유전자 알고리즘에 기반한 다목적 최적화 방법을 도입하였다. 또한, 최적화 과정에서 요구되는 비선형 시간이력해석을 수행하지 않고도, 비선형 이력감쇠기로 연결된 구조물의 지진응답을 효율적으로 평가하기 위하여 추계학적 선형화 방법을 접목하였다. 제시하는 방법의 효율성을 검증하기 위한 수치 예로서 20층과 10층의 인접 빌딩을 고려하였으며, 두 빌딩을 연결하는 비선형 감쇠시스템으로는 입력전압의 크기에 따라 변화하는 감쇠성능을 보이는 MR 감쇠기를 도입하였다. 제시하는 방법을 통하여 MR 감쇠기의 각 층별 최적 개수 및 감쇠용량을 결정할 수 있었으며, 이는 일반적인 균등분포 시스템에 비해 유사한 제어성능을 보이면서도 훨씬 경제적이었다. 또한, 인접구조물간 충돌에 대하여도 확률적으로 안정적인 거동을 보임을 검증하였으며, 제시하는 방법이 준능동 제어시스템의 최적 배치를 결정하기 위한 설계문제에도 적용할 수 있음을 보였다.

• Journal of Computational Design and Engineering
• /
• 제1권3호
• /
• pp.187-193
• /
• 2014

One of the hottest challenges in automotive industry is related to weight reduction in sheet metal forming processes, in order to produce a high quality metal part with minimal material cost. Stamping is the most widely used sheet metal forming process; but its implementation comes with several fabrication flaws such as springback and failure. A global and simple approach to circumvent these unwanted process drawbacks consists in optimizing the initial blank shape with innovative methods. The aim of this paper is to introduce an efficient methodology to deal with complex, computationally expensive multicriteria optimization problems. Our approach is based on the combination of methods to capture the Pareto Front, approximate criteria (to save computational costs) and global optimizers. To illustrate the efficiency, we consider the stamping of an industrial workpiece as test-case. Our approach is applied to the springback and failure criteria. To optimize these two criteria, a global optimization algorithm was chosen. It is the Simulated Annealing algorithm hybridized with the Simultaneous Perturbation Stochastic Approximation in order to gain in time and in precision. The multicriteria problems amounts to the capture of the Pareto Front associated to the two criteria. Normal Boundary Intersection and Normalized Normal Constraint Method are considered for generating a set of Pareto-optimal solutions with the characteristic of uniform distribution of front points. The computational results are compared to those obtained with the well-known Non-dominated Sorting Genetic Algorithm II. The results show that our proposed approach is efficient to deal with the multicriteria shape optimization of highly non-linear mechanical systems.

• 대한기계학회논문집B
• /
• 제34권7호
• /
• pp.679-688
• /
• 2010

본 연구에서는 기존의 역열전달 문제(inverse heat transfer problem)와 같이 역해석(inverse analysis)을 통해 미지의 파라미터를 추정(estimation)하는 개념을 복잡한 연소문제에 도입하였다. 기존의 연구에서는 역해석 기법을 연소문제 자체에 보다는 대부분 연소현상을 동반한 복사열전달과 같은 역열전달 문제에 국한해서 적용하고 있기 때문에, 열전달 문제에 한정되어 사용되고 있는 기존의 역해석을 새로운 공학문제에 확장하여 적용함과 동시에 효율적인 연소기 설계 및 최적화 개념을 제시하는데 본 연구의 의의가 있다고 할 수 있다. 이를 위해 실제적으로 많이 사용하고 있는 축대칭 원통형 연소기 내부로 주입되는 메탄($CH_4$)과 산소($O_2$) 성분의 초기 질량분율 값을 연소기 입구 근방에서 측정한 개스의 온도 데이터를 이용하여 역추정하였다. 이때, 복잡한 확산지배 연소 현상을 효율적으로 역해석하기 위해 최적화 방법 중의 하나인 반발 입자 군집 최적화 방법을 역해석 기법으로 적용하였다.

• Asian-Australasian Journal of Animal Sciences
• /
• 제21권11호
• /
• pp.1559-1571
• /
• 2008

A rule was developed to constrain the annual rate of inbreeding to a predefined value in a population with different lifetimes between sires and dams, and to maximize the selection response over generations. This rule considers that the animals in a population should be divided into sex-age classes based on the theory of gene flow, and restricts the increase of average inbreeding coefficient for new offspring by limiting the increase of the mean additive genetic relationship for parents selected. The optimization problem of this rule was formulated as a quadratic programming problem. Inputs for the rule were the BLUP estimated breeding values, the additive genetic relationship matrix of all animals, and the long-term contributions of sex-age classes. Outputs were optimal number and contributions of selected animals. In addition, this rule was combined with the optimization of emphasis given to QTL, and further increased the genetic gain over the planning horizon. Stochastic simulations of closed nucleus schemes for pigs were used to investigate the potential advantages obtained from this rule by combining the standard QTL selection, optimal QTL selection and conventional BLUP selection. Results showed that the predefined rates of inbreeding were actually achieved by this rule in three selection strategies. The rule obtained up to 9.23% extra genetic gain over truncation selection at the same rates of inbreeding. The combination of the extended rule and the optimization of emphasis given to QTL allowed substantial increases in selection response at a fixed annual rate of inbreeding, and solved substantially the conflict between short-term and long-term selection response in QTL-assisted selection schemes.