• Proceedings of the Korea Water Resources Association Conference
• /
• 2015.05a
• /
• pp.475-478
• /
• 2015

The objective of this study is to apply a hybrid model for estimating solar radiation and investigate their accuracy. A hybrid model is wavelet-based support vector machines (WSVMs). Wavelet decomposition is employed to decompose the solar radiation time series into approximation and detail components. These decomposed time series are then used as inputs of support vector machines (SVMs) modules in the WSVMs model. Results obtained indicate that WSVMs can successfully be used for the estimation of daily global solar radiation at Champaign and Springfield stations in Illinois.

• Communications of Mathematical Education
• /
• v.26 no.1
• /
• pp.137-154
• /
• 2012

Approximation' is one of central conceptions in calculus. A basic conception for explaining 'approximation' is 'tangent', and 'tangent' is a 'line' with special condition. In this study, we will study pedagogically these mathematical knowledge on the ground of a viewpoint on the teaching of secondary geometry, and in connection with these we will suggest the teaching program and the chief end for the probable teaching. For this, we will examine point, line, circle, straight line, tangent line, approximation, and drive meaningfully mathematical knowledge for algebraic operation through the process translating from the above into analytic geometry. And we will construct the stream line of mathematical knowledge for approximation from a view of modern mathematics. This study help mathematics teachers to promote the pedagogical content knowledge, and to provide the basis for development of teaching model guiding the mathematical knowledge. Moreover, this study help students to recognize that mathematics is a systematic discipline and school mathematics are activities constructed under a fixed purpose.

• The Journal of the Korea Contents Association
• /
• v.21 no.3
• /
• pp.616-625
• /
• 2021

Deep neural networks are an approximation method that approximates an arbitrary function to a linear model and then repeats additional approximation using a nonlinear active function. In this process, the method of evaluating the performance of approximation uses the loss function. Existing in-depth learning methods implement approximation that takes into account loss functions in the linear approximation process, but non-linear approximation phases that use active functions use non-linear transformation that is not related to reduction of loss functions of loss. This study proposes parametric activation functions that introduce scale parameters that can change the scale of activation functions and location parameters that can change the location of activation functions. By introducing parametric activation functions based on scale and location parameters, the performance of nonlinear approximation using activation functions can be improved. The scale and location parameters in each hidden layer can improve the performance of the deep neural network by determining parameters that minimize the loss function value through the learning process using the primary differential coefficient of the loss function for the parameters in the backpropagation. Through MNIST classification problems and XOR problems, parametric activation functions have been found to have superior performance over existing activation functions.

• Geophysics and Geophysical Exploration
• /
• v.1 no.2
• /
• pp.127-135
• /
• 1998

The Born approximation is widely used for solving the complex scattering problems in electromagnetics. Approximating total internal electric field by the background field is reasonable for small material contrasts as long as scatterer is not too large and the frequency is not too high. However in many geophysical applications, moderate and high conductivity contrasts cause both real and imaginary part of internal electric field to differ greatly from background. In the extended Born approximation, which can improve the accuracy of Born approximation dramatically, the total electric field in the integral over the scattering volume is approximated by the background electric field projected to a depolarization tensor. The finite difference and elements methods are usually used in EM scattering problems with a 2D model and a 3D source, due to their capability for simulating complex subsurface conductivity distributions. The price paid for a 3D source is that many wavenumber domain solutions and their inverse Fourier transform must be computed. In these differential equation methods, all the area including homogeneous region should be discretized, which increases the number of nodes and matrix size. Therefore, the differential equation methods need a lot of computing time and large memory. In this study, EM modeling program for a 2D model and a 3D source is developed, which is based on the extended Born approximation. The solution is very fast and stable. Using the program, crosshole EM responses with a vertical magnetic dipole source are obtained and the results are compared with those of 3D integral equation solutions. The agreement between the integral equation solution and extended Born approximation is remarkable within the entire frequency range, but degrades with the increase of conductivity contrast between anomalous body and background medium. The extended Born approximation is accurate in the case conductivity contrast is lower than 1:10. Therefore, the location and conductivity of the anomalous body can be estimated effectively by the extended Born approximation although the quantitative estimate of conductivity is difficult for the case conductivity contrast is too high.

• Tunnel and Underground Space
• /
• v.20 no.3
• /
• pp.153-168
• /
• 2010

A fifth equation is required to determine the five independent elastic constants of a transversely isotropic rock from compression test of a single specimen. As an approximation proposed by Saint-Venant has been used for long time, it may cause an erroneous result in some cases, especially for specimen with low angle of anisotropy. Three equations were suggested replacing this traditional equation and proved to be applicable by the model analysis in the previous studies. As Saint-Venant's approximation is turned out the same as the first one of them, it has the characteristics that the apparent Young's modulus is monotonously increasing according to the anisotropic angle. The methodology to analyze the elastic constants from four independent strain measurements by uniaxial compressive test of a single standard specimen is concisely described, and the necessity and compatibility of new suggested equations are discussed. Saint-Venant's approximation can determine the elastic constants close to true values and other equations may be unnecessary in specimens with medium to large angle. Nevertheless, they may become applicable because they can produce the almost same amount. For the specimens of small angle of anisotropy, Saint-Venant's approximation may result in out of general ranges or thermodynamic constraints, but other suggested equations can produce the almost true value. Thus they can be applied before other alternative equation is known. The guide map constructed by model study may decide the most compatible one of the three equations.

• Journal of Electrical Engineering and Technology
• /
• v.12 no.6
• /
• pp.2187-2195
• /
• 2017

In order to solve the AC optimal power flow (OPF) problem considering the generators' on/off status, it is necessary to model the problem as mixed-integer nonlinear programming (MINLP). Because the computation time to find the optimal solution to the mixed-integer AC OPF problem increases significantly as the system becomes larger, most of the existing solutions simplify the problem either by deciding the on/off status of generators using a separate unit commitment algorithm or by ignoring the minimum output of the generators. Even though this kind of simplification may make the overall computation time tractable, the results can be significantly erroneous. This paper proposes a novel algorithm for the mixed-integer AC OPF problem, which can provide a near-optimal solution quickly and efficiently. The proposed method is based on a combination of the outer approximation method and the relaxed AC OPF theory. The method is applied to a real-scale power system that has 457 generators and 2132 buses, and the result is compared to the branch-and-bound (B&B) method and the genetic algorithm. The results of the proposed method are almost identical to those of the compared methods, but computation time is significantly shorter.

• Nuclear Engineering and Technology
• /
• v.20 no.3
• /
• pp.197-202
• /
• 1988

A finite slab approximation method was developed to predict the long-term teachability. It is based on the assumption that the diffusional characteristics of radionuclides in a waste matrix are not dependent on matrix geometry but dependent on volume to surface ratio V/S) and diffusion coefficient. Consequently it can be expressed as the solution of the equations obtained from a finite slab with an equal V/S ratio (imaginary diffusion length). The calculational results by the finite slab approximation method have been compared with the results obtained for finite cylinder and sphere with corresponding diffusional analysis. The results of this simple model have showed a good agreement and presented a general applicability for the long-term prediction of the radionuclide leaching behavior.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.35C no.2
• /
• pp.56-64
• /
• 1998

Multi-layered perceptrons that are a nonlinear neural network model, have been widely used for various applications mainly thanks to good function approximation capability for nonlinear fuctions. However, for digital hardware implementation of the multi-layere perceptrons, the quantization scheme using "look-up tables (LUTs)" is commonly employed to handle nonlinear signmoid activation functions in the neworks, and thus requires large amount of storage to prevent unacceptable quantization errors. This paper is concerned with a new effective methodology for digital hardware implementation of multi-layered perceptrons, and proposes a "piecewise affine approximation" method in which input domain is divided into (small number of) sub-intervals and nonlinear sigmoid function is linearly approximated within each sub-interval. Using the proposed method, we develop an expression and an error backpropagation type learning algorithm for a multi-layered perceptron, and compare the performance with the quantization method through Monte Carlo simulations on XOR problems. Simulation results show that, in terms of learning convergece, the proposed method with a small number of sub-intervals significantly outperforms the quantization method with a very large storage requirement. We expect from these results that the proposed method can be utilized in digital system implementation to significantly reduce the storage requirement, quantization error, and learning time of the quantization method.quantization method.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.34 no.2
• /
• pp.191-201
• /
• 2010

This paper presents a new global optimization algorithm based on the branch-and-bound principle using Bspline approximation techniques. It describes the algorithmic components and details on their implementation. The key components include the subdivision of a design space into mutually disjoint subspaces and the bound calculation of the subspaces, which are all established by a real-valued B-spline volume model. The proposed approach was demonstrated with various test problems to reveal computational performances such as the solution accuracy, number of function evaluations, running time, memory usage, and algorithm convergence. The results showed that the proposed algorithm is complete without using heuristics and has a good possibility for application in large-scale NP-hard optimization.

• Journal of Institute of Control, Robotics and Systems
• /
• v.22 no.10
• /
• pp.779-787
• /
• 2016

A nonlinear control law for the quad-rotor of a low-complexity, global approximation-free from system uncertainties and external disturbances are described in this paper. The control law guarantees convergence to a small bounded error using a prescribed performance function. The stability of the proposed nonlinear control system is also proven by the Lyapunov stability theorem. The advantage of this technique is that it has a simpler form than any other nonlinear compensators and is applicable to any nonlinear systems without precise knowledge of the systems. In this paper, the proposed approach is applied to attitude/altitude control of a quad-rotor. Numerical simulations are performed to investigate the proposed nonlinear attitude control law by applying it to an uncertain quadcopter system with external disturbances.