• Proceedings of the KIEE Conference
• /
• 1998.07b
• /
• pp.740-743
• /
• 1998

This paper presents an optimization algorithm for a stable Dynamic Neural Network (DNN) using genetic algorithm. Optimized DNN is applied to a problem of controlling nonlinear dynamical systems. DNN is dynamic mapping and is better suited for dynamical systems than static forward neural network. The real time implementation is very important, and thus the neuro controller also needs to be designed such that it converges with a relatively small number of training cycles. SDNN has considerably fewer weights than DNN. The object of proposed algorithm is to the number of self dynamic neuron node and the gradient of activation functions are simultaneously optimized by genetic algorithms. To guarantee convergence, an analytic method based on the Lyapunov function is used to find a stable learning for the SDNN. The ability and effectiveness of identifying and controlling, a nonlinear dynamic system using the proposed optimized SDNN considering stability' is demonstrated by case studies.

• Nonlinear Functional Analysis and Applications
• /
• v.28 no.2
• /
• pp.497-520
• /
• 2023

Numerous problems in science and engineering defined by nonlinear functional equations can be solved by reducing them to an equivalent fixed point problem. Fixed point theory provides essential tools for solving problems arising in various branches of mathematical analysis, such as split feasibility problems, variational inequality problems, nonlinear optimization problems, equilibrium problems, complementarity problems, selection and matching problems, and problems of proving the existence of solution of integral and differential equations.The theory of fixed is known to find its applications in many fields of science and technology. For instance, the whole world has been profoundly impacted by the novel Coronavirus since 2019 and it is imperative to depict the spread of the coronavirus. Panda et al. [24] applied fractional derivatives to improve the 2019-nCoV/SARS-CoV-2 models, and by means of fixed point theory, existence and uniqueness of solutions of the models were proved. For more information on applications of fixed point theory to real life problems, authors should (see [6, 13, 24] and the references contained in).

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.42 no.4
• /
• pp.739-746
• /
• 2017

Orthogonal frequency division multiplexing(OFDM) systems have many advantages. However, OFDM systems are much affected by a nonlinear distortion because those systems have a high peak to average power ratio(PAPR) value. A selected mapping technology was suggested to reduce a PAPR value. The technology does not have data loss but receivers need side information to know modified phase sequence. Therefore, side information causes decreased a transmission efficiency. In this paper, we suggest a blind SLM receiver using a cross correlation technology. This receiver does not require side information. The proposed blind SLM receiver calculates sums of cross-correlation between transmitted pilot signals multiplied by each phase sequence and received pilot signals. So, this receiver detects side information which has a maximum sum cross-correlation value. We compared our proposed SLM receiver to a conventional blind SLM receiver through bit error rate(BER) and side information error rate(SIER) performances. Simulation results show that the proposed SLM receiver has improved BER and SIER performances than the conventional SLM receiver.

• The KIPS Transactions:PartA
• /
• v.14A no.1 s.105
• /
• pp.73-76
• /
• 2007

The determination of the conformal maps from the unit disk onto a Jordan region has been completed by solving the Theodorsen equation which is an nonlinear equation for the boundary correspondence function. Wegmann's method has been well known for the efficient mothed among the many suggestions for the Theodorsen equation. We proposed an improved method for convergence by applying a low-frequency pass filter to the Wegmann's method and theoretically proved convergence of improved iteration[1, 2]. And we proposed an effective method which makes it possible to estimate an error even if the real value is nut acquired[3]. In this paper, we propose an automatic algorithm for numerical conformal mapping bared on this error analysis in our early study. By this algorithm numerical conformal mapping is determined automatically according to the given domain of problem and the required accuracy. The discrete numbers and parameters of the low-frequency filter were acquired only by experience. This algorithm, however, is able to determine the discrete numbers and parameters of the low-frequency filter automatically in accordance with the given region This results from analyzing the function, which may decide the shape of the given domain under the assumption that the degree of the problem depends of the transformation of a given domain, as seen in the Fourier Transform. This proposed algorithm is also ploved by numerical experience.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.2
• /
• pp.1-10
• /
• 2002

This paper presents the new template based human face recognition methods to improve the recognition performance against scale and in-plane rotation variations of face images. To enhance the recognition performance, the templates are generated by linear or nonlinear operation on multiple images including different scales and rotations of faces. As the invariant features to allow for scale and rotation variations of face images, we adopt the affine transformation, the log-polar mapping, and the log-polar image based FFT. The proposed recognition methods are evaluated in terms of the recognition rate and the processing time. Experimental results show that the proposed template based methods lead to higher recognition rate than the single image based one. The affine transformation based face recognition method shows marginally higher recognition rate than those of the log-polar mapping based method and the log-polar image based FFT, while, in the aspect of processing time, the log-polar mapping based method is the fastest one.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.10
• /
• pp.75-80
• /
• 2017

In this paper, we describe the effect of the regularization method and the network with identity mapping on the performance of the activation functions in deep convolutional neural networks. The activation functions act as nonlinear transformation. In early convolutional neural networks, a sigmoid function was used. To overcome the problem of the existing activation functions such as gradient vanishing, various activation functions were developed such as ReLU, Leaky ReLU, parametric ReLU, and ELU. To solve the overfitting problem, regularization methods such as dropout and batch normalization were developed on the sidelines of the activation functions. Additionally, data augmentation is usually applied to deep learning to avoid overfitting. The activation functions mentioned above have different characteristics, but the new regularization method and the network with identity mapping were validated only using ReLU. Therefore, we have experimentally shown the effect of the regularization method and the network with identity mapping on the performance of the activation functions. Through this analysis, we have presented the tendency of the performance of activation functions according to regularization and identity mapping. These results will reduce the number of training trials to find the best activation function.

• The Journal of the Korea Contents Association
• /
• v.5 no.1
• /
• pp.201-208
• /
• 2005

Tn this paper, we evaluate the peak-to-average power ratio (PAPR) performance in a space-time block code (STBC) multi-input multi-output orthogonal frequency division multiplexing (MIMO-OFDM) system using selected mapping (SLM) and partial transmit sequences (PTS) approaches. SLM and PTS methods are used to decrease the nonlinear distortion and to improve the power efficiency of the nonlinear high power amplifier(HPA) in the MIMO-OFDM system. In simulation result, when compared with the existing MIMO-OFDM system using QPSK, the PTS method reduces the PAPR about 5dB while the SLM method can reduce about 3.5 dB. Also, we find the BER performance of the MIMO-OFDM system with and without the predistorter in front of the HPA. When the predistorter is used, the input back-off (IBO) of 4 dB is required in the PTS method, and IBO of 6 dB in the SLM method to closely conform to the linear amplifier. If the method of improving the PAPR is not used, the value of IBO of 8 dB is required.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.29 no.3
• /
• pp.237-244
• /
• 2016

This paper presents a nonlinear Moving Least Squares(MLS) difference method for material nonlinearity problem. The MLS difference method, which employs strong formulation involving the fast derivative approximation, discretizes governing partial differential equation based on a node model. However, the conventional MLS difference method cannot explicitly handle constitutive equation since it solves solid mechanics problems by using the Navier's equation that unifies unknowns into one variable, displacement. In this study, a double derivative approximation is devised to treat the constitutive equation of inelastic material in the framework of strong formulation; in fact, it manipulates the first order derivative approximation two times. The equilibrium equation described by the divergence of stress tensor is directly discretized and is linearized by the Newton method; as a result, an iterative procedure is developed to find convergent solution. Stresses and internal variables are calculated and updated by the return mapping algorithm. Effectiveness and stability of the iterative procedure is improved by using algorithmic tangent modulus. The consistency of the double derivative approximation was shown by the reproducing property test. Also, accuracy and stability of the procedure were verified by analyzing inelastic beam under incremental tensile loading.

• Journal of Information Processing Systems
• /
• v.17 no.5
• /
• pp.1020-1033
• /
• 2021

Nowadays, online or mobile social network services (SNS) are very popular and widely spread in our society and daily lives to instantly share, disseminate, and search information. In particular, SNS such as YouTube, Flickr, Facebook, and Amazon allow users to upload billions of images or videos and also provide a number of multimedia information to users. Information retrieval in multimedia-rich SNS is very useful but challenging task. Content-based media retrieval (CBMR) is the process of obtaining the relevant image or video objects for a given query from a collection of information sources. However, CBMR suffers from the dimensionality curse due to inherent high dimensionality features of media data. This paper investigates the effectiveness of the kernel trick in CBMR, specifically, the kernel principal component analysis (KPCA) for dimensionality reduction. KPCA is a nonlinear extension of linear principal component analysis (LPCA) to discovering nonlinear embeddings using the kernel trick. The fundamental idea of KPCA is mapping the input data into a highdimensional feature space through a nonlinear kernel function and then computing the principal components on that mapped space. This paper investigates the potential of KPCA in CBMR for feature extraction or dimensionality reduction. Using the Gaussian kernel in our experiments, we compute the principal components of an image dataset in the transformed space and then we use them as new feature dimensions for the image dataset. Moreover, KPCA can be applied to other many domains including CBMR, where LPCA has been used to extract features and where the nonlinear extension would be effective. Our results from extensive experiments demonstrate that the potential of KPCA is very encouraging compared with LPCA in CBMR.

• Journal of Sensor Science and Technology
• /
• v.32 no.2
• /
• pp.110-117
• /
• 2023

Reconstructing underwater geometry in real time with forward-looking sonar is critical for applications such as localization, mapping, and path planning. Geometrical data must be repeatedly calculated and overwritten in real time because the reliability of the acoustic data is affected by various factors. Moreover, scattering of signal data during the coordinate conversion process may lead to geometrical errors, which lowers the accuracy of the information obtained by the sensor system. In this study, we propose a three-step data processing method with low computational cost for real-time operation. First, the number of data points to be interpolated is determined with respect to the distance between each point and the size of the data grid in a Cartesian coordinate system. Then, the data are processed with a nonlinear interpolation so that they exhibit linear properties in the coordinate system. Finally, the data are transformed based on variations in the position and orientation of the sonar over time. The results of an evaluation of our proposed approach in a simulation show that the nonlinear interpolation operation constructed a continuous underwater geometry dataset with low geometrical error.