• Proceedings of the KIEE Conference
• /
• 2008.07a
• /
• pp.1857-1858
• /
• 2008

본 논문에서는 FCM 기반 RBF 뉴럴네트워크(FCM-RBFNN) 구조를 제안하고 PSO를 이용한 FCM-RBFNN의 구조 및 파라미터의 최적화 방법을 제시한다. 클러스터링 알고리즘은 퍼지 뉴럴 네트워크에서 멤버쉽함수의 중심점과 반경 등을 결정하는 학습에 일반적으로 사용된다. 제안된 FCM-RBFNN서는 방사기저함수로써 가우시안, 삼각형 타입 등의 정해진 형태를 사용하지 않고 데이터들 사이의 거리에 관계된 계산을 수행하는 FCM에 의해 결정된다. 기존의 RBFNN에서 후반부는 상수형태로써 방사기저함수의 선형결합으로써 표현되는 반면에 제안된 FCM-RBFNN의 후반부는 상수형, 선형, 2차식 등의 다양한 형태의 다항식으로 표현될 수 있으며 다항식의 계수는 WLSE를 이용하여 추정한다. FCM 기반 RBF 뉴럴 네트워크의 성능은 퍼지규칙의 수, 후반부 다항식의 차수 FCM의 퍼지화 계수에 의하여 결정기 때문에 FCM-RBFNN의 구조와 파라미터의 최적화가 요구된다. 본 논문에서는 PSO를 이용하여 FCM-RBFNN의 구조에 관련된 퍼지 규칙의 수, 후반부 다항식의 차수와 파라미터에 관련된 퍼지화 계수를 최적화한다. 또한 후반부 다항식의 계수는 WLSE를 사용하여 추정한다.

• Computers and Concrete
• /
• v.16 no.5
• /
• pp.741-757
• /
• 2015

In the last decade, several modeling approaches have been proposed and applied to estimate the high-performance concrete (HPC) slump flow. While HPC is a highly complex material, modeling its behavior is a very difficult issue. Thus, the selection and application of proper modeling methods remain therefore a crucial task. Like many other applications, HPC slump flow prediction suffers from noise which negatively affects the prediction accuracy and increases the variance. In the recent years, ensemble learning methods have introduced to optimize the prediction accuracy and reduce the prediction error. This study investigates the potential usage of bagging (Bag), which is among the most popular ensemble learning methods, in building ensemble models. Four well-known artificial intelligence models (i.e., classification and regression trees CART, support vector machines SVM, multilayer perceptron MLP and radial basis function neural networks RBF) are deployed as base learner. As a result of this study, bagging ensemble models (i.e., Bag-SVM, Bag-RT, Bag-MLP and Bag-RBF) are found superior to their base learners (i.e., SVM, CART, MLP and RBF) and bagging could noticeable optimize prediction accuracy and reduce the prediction error of proposed predictive models.

• Advances in concrete construction
• /
• v.13 no.1
• /
• pp.11-23
• /
• 2022

Conventional concrete needs some improvement in the mechanical properties, which can be obtained by different admixtures. However, making concrete samples costume always time and money. In this paper, different types of hybrid algorithms are applied to develop predictive models for forecasting compressive strength (CS) of concretes containing metakaolin (MK) and fly ash (FA). In this regard, three different algorithms have been used, namely multilayer perceptron (MLP), radial basis function (RBF), and support vector machine (SVR), to predict CS of concretes by considering most influencers input variables. These algorithms integrated with the grey wolf optimization (GWO) algorithm to increase the model's accuracy in predicting (GWMLP, GWRBF, and GWSVR). The proposed MLP models were implemented and evaluated in three different layers, wherein each layer, GWO, fitted the best neuron number of the hidden layer. Correspondingly, the key parameters of the SVR model are identified using the GWO method. Also, the optimization algorithm determines the hidden neurons' number and the spread value to set the RBF structure. The results show that the developed models all provide accurate predictions of the CS of concrete incorporating MK and FA with R² larger than 0.9972 and 0.9976 in the learning and testing stage, respectively. Regarding GWMLP models, the GWMLP1 model outperforms other GWMLP networks. All in all, GWSVR has the worst performance with the lowest indices, while the highest score belongs to GWRBF.

• Advances in environmental research
• /
• v.5 no.3
• /
• pp.153-167
• /
• 2016

We applied multilayer perceptron (MLP) and radial basis function (RBF) neural network in upstream and downstream water quality stations of the Karaj Reservoir in Iran. For both neural networks, inputs were pH, turbidity, temperature, chlorophyll-a, biochemical oxygen demand (BOD) and nitrate, and the output was dissolved oxygen (DO). We used an MLP neural network with two hidden layers, for upstream station 15 and 33 neurons in the first and second layers respectively, and for the downstream station, 16 and 21 neurons in the first and second hidden layer were used which had minimum amount of errors. For learning process 6-fold cross validation were applied to avoid over fitting. The best results acquired from RBF model, in which the mean bias error (MBE) and root mean squared error (RMSE) were 0.063 and 0.10 for the upstream station. The MBE and RSME were 0.0126 and 0.099 for the downstream station. The coefficient of determination ($R^2$) between the observed data and the predicted data for upstream and downstream stations in the MLP was 0.801 and 0.904, respectively, and in the RBF network were 0.962 and 0.97, respectively. The MLP neural network had acceptable results; however, the results of RBF network were more accurate. A sensitivity analysis for the MLP neural network indicated that temperature was the first parameter, pH the second and nitrate was the last factor affecting the prediction of DO concentrations. The results proved the workability and accuracy of the RBF model in the prediction of the DO.

• Journal of Korea Multimedia Society
• /
• v.13 no.5
• /
• pp.722-729
• /
• 2010

Prediction problems are widely used in medical domains. For example, computer aided diagnosis or prognosis is a key component in a CDSS (Clinical Decision Support System). SVMs with nonlinear kernels like RBF kernels, have shown superior accuracy in prediction problems. However, they are not preferred by physicians for medical prediction problems because nonlinear SVMs are difficult to visualize, thus it is hard to provide intuitive interpretation of prediction results to physicians. Nomogram was proposed to visualize SVM classification models. However, it cannot visualize nonlinear SVM models. Localized Radial Basis Function (LRBF) was proposed which shows comparable accuracy as the RBF kernel while the LRBF kernel is easier to interpret since it can be linearly decomposed. This paper presents a new tool named VRIFA, which integrates the nomogram and LRBF kernel to provide users with an interactive visualization of nonlinear SVM models, VRIFA visualizes the internal structure of nonlinear SVM models showing the effect of each feature, the magnitude of the effect, and the change at the prediction output. VRIFA also performs nomogram-based feature selection while training a model in order to remove noise or redundant features and improve the prediction accuracy. The area under the ROC curve (AUC) can be used to evaluate the prediction result when the data set is highly imbalanced. The tool can be used by biomedical researchers for computer-aided diagnosis and risk factor analysis for diseases.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.6
• /
• pp.1353-1366
• /
• 2015

These days Shewhart control chart for evaluating stability of the process is widely used in various field. But it must follow strict assumption of distribution. In real-life problems, this assumption is often violated when many quality characteristics follow non-normal distribution. Moreover, it is more serious in multivariate quality characteristics. To overcome this problem, many researchers have studied the non-parametric control charts. Recently, SVDD (Support Vector Data Description) control chart based on RBF (Radial Basis Function) Kernel, which is called K-chart, determines description of data region on in-control process and is used in various field. But it is important to select kernel parameter or etc. in order to apply the K-chart and they must be predetermined. For this, many researchers use grid search for optimizing parameters. But it has some problems such as selecting search range, calculating cost and time, etc. In this paper, we research the efficiency of selecting parameter regions as data structure vary via simulation study and propose a new method for determining parameters so that it can be easily used and discuss a robust choice of parameters for various data structures. In addition, we apply it on the real example and evaluate its performance.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2007.10a
• /
• pp.407-412
• /
• 2007

본 논문에서는 비선형 블라인드 채널등화기의 구현을 위하여 가우시안 가중치(gaussian weights)를 이용한 개선된 퍼지 클러스터(Modified Fuzzy C-Means with Gaussian Weights: MFCM_GW) 알고리즘을 제안한다. 제안된 알고리즘은 기존 FCM 알고리즘의 유클리디언 거리(Euclidean distance) 값 대신 Bayesian Likelihood 목적함수(fitness function)와 가우시안 가중치가 적용된 멤버쉽 매트릭스(partition matrix)를 이용하여, 비선형 채널의 출력으로 수신된 데이터들로부터 최적의 채널 출력 상태 값(optimal channel output states)들을 직접 추정한다. 이렇게 추정된 채널 출력 상태 값들로 비선형 채널의 이상적 채널 상태(desired channel states) 벡터들을 구성하고, 이를 Radial Basis Function(RBF) 등화기의 중심(center)으로 활용함으로써 송신된 데이터 심볼을 찾아낸다. 실험에서는 무작위 이진 신호에 가우시안 잡음이 추가된 데이터를 사용하여 기존의 Simplex Genetic Algorithm(GA), 하이브리드 형태의 GASA(GA merged with simulated annealing (SA)), 그리고 과거에 발표되었던 MFCM 등과 그 성능을 비교 분석하였으며, 가우시안 가중치가 적용된 MFCM_GW를 이용한 채널등화기가 상대적으로 정확도와 속도 면에서 우수함을 보였다.

• Journal of the Korean Society for Precision Engineering
• /
• v.24 no.11
• /
• pp.144-153
• /
• 2007

In order to generate the input data for rapid prototyping, a new approach which is based on the implicit surface interpolation method is presented. In the method a surface is reconstructed by creating smooth implicit surface from unorganized cloud of points through which the surface should pass. In the method an implicit surface is defined by the adaptive local shape functions including quadratic polynomial function, cubic polynomial function and RBF(Radial Basis Function). By the reconstruction of a surface, various types of error in raw STL file including degenerated triangles, undesirable holes with complex shapes and overlaps between triangles can be eliminated automatically. In order to get the slicing data for rapid prototyping an efficient intersection algorithm between implicit surface and plane is developed. For the direct usage for rapid prototyping, a robust transformation algorithm for the generation of complete STL data of solid type is also suggested.

• Environmental Engineering Research
• /
• v.21 no.4
• /
• pp.333-340
• /
• 2016

An Artificial Neural Network including a Radial Basis Function (RBF) and a Time Delay Neural Network (TDNN) was used to predict total dissolved solid (TDS) in the river Zayanderud. Water quality parameters in the river for ten years, 2001-2010, were prepared from data monitored by the Isfahan Regional Water Authority. A factor analysis was applied to select the inputs of water quality parameters, which obtained total hardness, bicarbonate, chloride and calcium. Input data to the neural networks were pH, $Na^+$, $Mg^{2+}$, Carbonate ($CO{_3}^{-2}$), $HCO{_3}^{-1}$, $Cl^-$, $Ca^{2+}$ and Total hardness. For learning process 5-fold cross validation were applied. In the best situation, the TDNN contained 2 hidden layers of 15 neurons in each of the layers and the RBF had one hidden layer with 100 neurons. The Mean Squared Error and the Mean Bias Error for the TDNN during the training process were 0.0006 and 0.0603 and for the RBF neural network the mentioned errors were 0.0001 and 0.0006, respectively. In the RBF, the coefficient of determination ($R^2$) and the index of agreement (IA) between the observed data and predicted data were 0.997 and 0.999, respectively. In the TDNN, the $R^2$ and the IA between the actual and predicted data were 0.957 and 0.985, respectively. The results of sensitivity illustrated that $Ca^{2+}$ and $SO{_4}^{2-}$ parameters had the highest effect on the TDS prediction.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.6 no.3
• /
• pp.25-36
• /
• 1996

Conventional radial-basis-function networks and fuzzy systems have serious problems in dealing with the non1inea:r approximations on high-dimensional spaces due to the explosive increase of the number of hidden nodes or fuzzy IF-THEN rules. In order to avoid such problems, this paper proposes a tree-structured intelligent system in which semi-local basis functions form its basic elements, and develops a training algorithm for the proposed system based on the modified genetic algorithm and LMS rule. Theoretical analysis is performed on the approximation capability of the proposed system, together with experimental studies which demonstrate the effectiveness of the developed methodology.