• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2015년도 학술발표회
• /
• pp.606-609
• /
• 2015

본 연구에서는 support vector regression (SVR) 및 매개변수 최적화 알고리즘을 이용한 하천수위 예측모델을 구축하고 이를 실제 유역에 적용하여 모델 효율성을 평가하였다. 여기서, SVR은 하천수위를 예측하기 위한 예측모델로서 채택되었으며, 커널함수 (Kernel function)로서는 radial basis function (RBF)을 선택하였다. 최적화 알고리즘은 SVR의 최적 매개변수 (C?, cost parameter or regularization parameter; ${\gamma}$, RBF parameter; ${\epsilon}$, insensitive loss function parameter)를 탐색하기 위하여 적용되었다. 매개변수 최적화 알고리즘으로는 grid search (GS), genetic algorithm (GA), particle swarm optimization (PSO), artificial bee colony (ABC) 알고리즘을 채택하였으며, 비교분석을 통해 최적화 알고리즘의 적용성을 평가하였다. 또한 SVR과 최적화 알고리즘을 결합한 모델 (SVR-GS, SVR-GA, SVR-PSO, SVR-ABC)은 기존에 수자원 분야에서 널리 적용되어온 신경망(Artificial neural network, ANN) 및 뉴로퍼지 (Adaptive neuro-fuzzy inference system, ANFIS) 모델과 비교하였다. 그 결과, 모델 효율성 측면에서 SVR-GS, SVR-GA, SVR-PSO 및 SVR-ABC는 ANN보다 우수한 결과를 나타내었으며, ANFIS와는 비슷한 결과를 나타내었다. 또한 SVR-GA, SVR-PSO 및 SVR-ABC는 SVR-GS보다 상대적으로 우수한 결과를 나타내었으며, 모델 효율성 측면에서 SVR-PSO 및 SVR-ABC는 가장 우수한 모델 성능을 나타내었다. 따라서 본 연구에서 적용한 매개변수 최적화 알고리즘은 SVR의 매개변수를 최적화하는데 효과적임을 확인할 수 있었다. SVR과 최적화 알고리즘을 이용한 하천수위 예측모델은 기존의 ANN 및 ANFIS 모델과 더불어 하천수위 예측을 위한 효과적인 도구로 사용될 수 있을 것으로 판단된다.

• 지구물리와물리탐사
• /
• 제1권1호
• /
• pp.64-70
• /
• 1998

회절지오토모그래피는 정량적으로 고 분해능의 지하구조를 구하는 영상화 기법이다. 일반적으로 회절지오토모그래피는 시추공-시추공 자료의 영상단면을 구하는데 적용된다. 본 연구에서는 일반적인 지표레이다 탐사자료로부터 고 분해능의 지하구조를 영상화하는 회절지오토모그래피 알고리듬을 구성하였다. 개발된 알고리듬을 유한차분 모형응답자료에 적용하여 적용조건, 적용한계 등을 살펴보았다. 역산 매개변수들(영상화 주파수의 개수, 평활화 계수, 주파수 영역)을 약산란 가정에 부합하는 고립형 대상체로부터 추정하였다. 또한 약산란 가정에 위배되는 모형에 적용하여 알고리듬의 유용성을 확인하였다.

• Smart Structures and Systems
• /
• 제23권3호
• /
• pp.295-306
• /
• 2019

Shanghai Tower is a 632-meter super high-rise building located in an area with wind and active earthquake. A sophisticated structural health monitoring (SHM) system consisting of more than 400 sensors has been built to carry out a long-term monitoring for its operational safety. In this paper, a reduced-order model including 31 elements was generated from a full model of this super tall building. An iterative regularized matrix method was proposed to tune the system parameters, making the dynamic characteristic of the reduced-order model be consistent with those in the full model. The updating reduced-order model can be regarded as a benchmark model for further analysis. A long-term monitoring for structural dynamic characteristics of Shanghai Tower under different construction stages was also investigated. The identified results, including natural frequency and damping ratio, were discussed. Based on the data collected from the SHM system, the dynamic characteristics of the whole structure was investigated. Compared with the result of the finite element model, a good agreement can be observed. The result provides a valuable reference for examining the evolution of future dynamic characteristics of this super tall building.

• 통합자연과학논문집
• /
• 제12권2호
• /
• pp.35-43
• /
• 2019

In the content-based image search properties, form information is simple because only the contours of objects are available, and although it can effectively extract the characteristics of the objects, it is sensitive to external noise. The radial distortion, one of these noises, is most prominent in the eyewear and, due to the structural characteristics of the imaging equipment, radiative distortion occurs in almost all imaging equipment. It is very important to determine the similarity of the objects in the images in which these distortions occurred to the actual objects. In order to improve this problem, we propose a strong image search technique for formative noise and radiative distortion using regularization phase angles and moments. Through simulation using Wang DB, the proposed algorithm proved excellent performance for radiation distortion that occurs in general. In addition, a system optimized for database can be implemented by making appropriate changes to the threshold values, enabling image retrieval with the desired level of confidence in various systems. The algorithm proposed in this paper is expected to be utilized as an optimal imaging system by extracting morphological form information of multimedia data.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권1호
• /
• pp.305-326
• /
• 2019

Discriminative correlation filter (DCF) based tracking algorithms have recently shown impressive performance on benchmark datasets. However, amount of recent researches are vulnerable to heavy occlusions, irregular deformations and so on. In this paper, we intend to solve these problems and handle the contradiction between accuracy and real-time in the framework of tracking-by-detection. Firstly, we propose an innovative strategy to combine the template and color-based models instead of a simple linear superposition and rely on the strengths of both to promote the accuracy. Secondly, to enhance the discriminative power of the learned template model, the spatial regularization is introduced in the learning stage to penalize the objective boundary information corresponding to features in the background. Thirdly, we utilize a discriminative multi-scale estimate method to solve the problem of scale variations. Finally, we research strategies to limit the computational complexity of our tracker. Abundant experiments demonstrate that our tracker performs superiorly against several advanced algorithms on both the OTB2013 and OTB2015 datasets while maintaining the high frame rates.

• Current Optics and Photonics
• /
• 제5권3호
• /
• pp.322-328
• /
• 2021

In this paper, a pretreatment method for a matrix in convex optimization is proposed to optimize the spectral reconstruction process of a disordered dispersion spectrometer. Unlike the reconstruction process of traditional spectrometers using Fourier transforms, the reconstruction process of disordered dispersion spectrometers involves solving a large-scale matrix equation. However, since the matrices in the matrix equation are obtained through measurement, they contain uncertainties due to out of band signals, background noise, rounding errors, temperature variations and so on. It is difficult to solve such a matrix equation by using ordinary nonstationary iterative methods, owing to instability problems. Although the smoothing Tikhonov regularization approach has the ability to approximatively solve the matrix equation and reconstruct most simple spectral shapes, it still suffers the limitations of reconstructing complex and irregular spectral shapes that are commonly used to distinguish different elements of detected targets with mixed substances by characteristic spectral peaks. Therefore, we propose a special pretreatment method for a matrix in convex optimization, which has been proved to be useful for reducing the condition number of matrices in the equation. In comparison with the reconstructed spectra gotten by the previous ordinary iterative method, the spectra obtained by the pretreatment method show obvious accuracy.

• Advances in nano research
• /
• 제12권5호
• /
• pp.489-499
• /
• 2022

Nanofluids have recently triggered a substantial scientific interest as cooling media. However, their stability is challenging for successful engagement in industrial applications. Different factors, including temperature, nanoparticles and base fluids characteristics, pH, ultrasonic power and frequency, agitation time, and surfactant type and concentration, determine the nanofluid stability regime. Indeed, it is often too complicated and even impossible to accurately find the conditions resulting in a stabilized nanofluid. Furthermore, there are no empirical, semi-empirical, and even intelligent scenarios for anticipating the stability of nanofluids. Therefore, this study introduces a straightforward and reliable intelligent classifier for discriminating among the stability regimes of alumina-water nanofluids based on the Zeta potential margins. In this regard, various intelligent classifiers (i.e., deep learning and multilayer perceptron neural network, decision tree, GoogleNet, and multi-output least squares support vector regression) have been designed, and their classification accuracy was compared. This comparison approved that the multilayer perceptron neural network (MLPNN) with the SoftMax activation function trained by the Bayesian regularization algorithm is the best classifier for the considered task. This intelligent classifier accurately detects the stability regimes of more than 90% of 345 different nanofluid samples. The overall classification accuracy and misclassification percent of 90.1% and 9.9% have been achieved by this model. This research is the first try toward anticipting the stability of water-alumin nanofluids from some easily measured independent variables.

• Steel and Composite Structures
• /
• 제44권1호
• /
• pp.119-139
• /
• 2022

This paper aims to use the artificial intelligence approach to develop a new model for predicting the ultimate axial strength of the circular concrete-filled steel tubular (CFST) stub columns. For this, the results of 314 experimentally tested circular CFST stub columns were employed in the generation of the design model. Since the influence of the column diameter, steel tube thickness, concrete compressive strength, steel tube yield strength, and column length on the ultimate axial strengths of columns were investigated in these experimental studies, here, in the development of the design model, these variables were taken into account as input parameters. The model was developed using the backpropagation algorithm named Bayesian Regularization. The accuracy, reliability, and consistency of the developed model were evaluated statistically, and also the design formulae given in the codes (EC4, ACI, AS, AIJ, and AISC) and the previous empirical formulations proposed by other researchers were used for the validation and comparison purposes. Based on this evaluation, it can be expressed that the developed design model has a strong and reliable prediction performance with a considerably high coefficient of determination (R-squared) value of 0.9994 and a low average percent error of 4.61. Besides, the sensitivity of the developed model was also monitored in terms of dimensional properties of columns and mechanical characteristics of materials. As a consequence, it can be stated that for the design of the ultimate axial capacity of the circular CFST stub columns, a novel artificial intelligence-based design model with a good and robust prediction performance was proposed herein.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권11호
• /
• pp.3913-3934
• /
• 2021

Real-time prediction of Web service of quality (QoS) provides more convenience for web services in cloud environment, but real-time QoS prediction faces severe challenges, especially under the cold-start situation. Existing literatures of real-time QoS predicting ignore that the QoS of a user/service is related to the QoS of other users/services. For example, users/services belonging to the same group of category will have similar QoS values. All of the methods ignore the group relationship because of the complexity of the model. Based on this, we propose a real-time Matrix Factorization based Clustering model (MFC), which uses category information as a new regularization term of the loss function. Specifically, in order to meet the real-time characteristic of the real-time prediction model, and to minimize the complexity of the model, we first map the QoS values of a large number of users/services to a lower-dimensional space by the PCA method, and then use the K-means algorithm calculates user/service category information, and use the average result to obtain a stable final clustering result. Extensive experiments on real-word datasets demonstrate that MFC outperforms other state-of-the-art prediction algorithms.

• International Journal of Internet, Broadcasting and Communication
• /
• 제14권1호
• /
• pp.136-141
• /
• 2022

A very complex task in deep learning such as image classification must be solved with the help of neural networks and activation functions. The backpropagation algorithm advances backward from the output layer towards the input layer, the gradients often get smaller and smaller and approach zero which eventually leaves the weights of the initial or lower layers nearly unchanged, as a result, the gradient descent never converges to the optimum. We propose a two-factor non-saturating activation functions known as Bea-Mish for machine learning applications in deep neural networks. Our method uses two factors, beta (𝛽) and alpha (𝛼), to normalize the area below the boundary in the Mish activation function and we regard these elements as Bea. Bea-Mish provide a clear understanding of the behaviors and conditions governing this regularization term can lead to a more principled approach for constructing better performing activation functions. We evaluate Bea-Mish results against Mish and Swish activation functions in various models and data sets. Empirical results show that our approach (Bea-Mish) outperforms native Mish using SqueezeNet backbone with an average precision (AP50val) of 2.51% in CIFAR-10 and top-1accuracy in ResNet-50 on ImageNet-1k. shows an improvement of 1.20%.