• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.11a
• /
• pp.427-430
• /
• 2007

기존의 LVQ(Learning Vector Quantization) 방법을 이용하여 물체를 분류하면 데이터의 학습이 빠르고 연산량이 적어 실시간으로 물체를 분류할 수 있는 장점이 있다. 하지만 데이터의 훈련시 output neuron의 개수를 정확히 예측할 수 없고 output neuron의 개수에 따라 물체를 분류하는 정확도가 매우 달라질 수 있다. 그러므로 본 논문에서는 output neuron의 개수를 데이터의 특성에 맞게 결정해주는 알고리즘을 제시한다. DLVQ(Dynamic Learning Vector Quantization) 알고리즘은 승자로 결정된 가중치 벡터의 부류가 샘플 데이터의 부류와 같으면 업데이트하고 다르면 새로운 가중치 벡터로 생성한다. 제한한 알고리즘의 가장 다른 부분은 미리 output neuron의 개수를 정하는 것이 아니라 훈련 과정에서 동적으로 output neuron의 개수를 생성하는 것이다. 그리고 클러터의 구분 방법을 제시하여 사람, 차, 클러터를 구분할 수 있다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.11
• /
• pp.5360-5376
• /
• 2017

In this paper, we study the cross-layer design for the AMC/HARQ-based wireless networks, and propose a new dynamic transmission-mode selection scheme to improve system spectrum efficiency. In the proposed scheme, dynamic thresholds for transmission-mode selection in each packet transmission and retransmission are jointly designed under the constraint of the overall packet error rate. Comparing with the existing schemes, the proposed scheme is inclined to apply higher modulation order at the first several (re)transmissions, which corresponds to higher-rate transmission modes thus higher average system spectrum efficiency. We also extend the cross-layer design to MIMO (Multi-input Multi-output) communication scenarios. Numerical results show that the proposed new dynamic transmission-mode selection scheme generally achieves higher average spectrum efficiency than the conventional and existing cross-layer design.

• Proceedings of the KSME Conference
• /
• 2008.11b
• /
• pp.3140-3141
• /
• 2008

The output power efficiency of the fuel cell system depends on the demanded current, stack temperature, air excess ratio, hydrogen excess ratio and inlet air humidity. Thus, it is necessary to determine the optimal operation condition for maximum power efficiency. In this paper, we developed a dynamic model of fuel cell system which contains mass flow model, diffusivity gas layer model, membrane hydration and electrochemistry model. In order to determine the maximum output power and minimum use of hydrogen in a certain power condition, response surface methodology (RSM) optimization based on the proposed PEMFC stack model is presented. The results provide an effective method to optimize the operation condition under varied situations.

• Journal of Magnetics
• /
• v.19 no.4
• /
• pp.315-322
• /
• 2014

In this work, Artificial Neural Network (ANN) was used to model the dynamic behavior of ferromagnetic hysteresis derived from performing the mean-field analysis on the Ising model. The effect of field parameters and system structure (via coordination number) on dynamic critical points was elucidated. The Ising magnetization equation was drawn from mean-field picture where the steady hysteresis loops were extracted, and series of the dynamic critical points for constructing dynamic phase-diagram were depicted. From the dynamic critical points, the field parameters and the coordination number were treated as inputs whereas the dynamic critical temperature was considered as the output of the ANN. The input-output datasets were divided into training, validating and testing datasets. The number of neurons in hidden layer was varied in structuring ANN network with highest accuracy. The network was then used to predict dynamic critical points of the untrained input. The predicted and the targeted outputs were found to match well over an extensive range even for systems with different structures and field parameters. This therefore confirms the ANN capabilities and indicates the ANN ability in modeling the ferromagnetic dynamic hysteresis behavior for establishing the dynamic-phase-diagram.

• Journal of the Korean Society for Precision Engineering
• /
• v.17 no.3
• /
• pp.90-101
• /
• 2000

In this paper, we propose a controller for differentially steered wheeled mobile robots. The controller uses input-output linearization algorithm and artificial neural network to stabilize the dynamic model and compensate uncertainties. The proposed neural network part has 6 inputs, 1 hidden layer, 2 torque outputs and features fast online learning and good performance on structure error learning basis. Simulation results show that the proposed controller perform precisely tracking of reference path and is robust to uncertainties.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.2
• /
• pp.159-167
• /
• 1996

In this paper, a new competitive learning algorithm called dynamic competitive learning (DCL) is presented. DCL is a supervised learning mehtod that dynamically generates output neuraons and nitializes weight vectors from training patterns. It introduces a new parameter called LOG (limit of garde) to decide whether or not an output neuron is created. In other words, if there exist some neurons in the province of LOG that classify the input vector correctly, then DCL adjusts the weight vector for the neuraon which has the minimum grade. Otherwise, it produces a new output neuron using the given input vector. It is largely learning is not limited only to the winner and the output neurons are dynamically generated int he trining process. In addition, the proposed algorithm has a small number of parameters. Which are easy to be determined and applied to the real problems. Experimental results for patterns recognition of remote sensing data and handwritten numeral data indicate the superiority of dCL in comparison to the conventional competitive learning methods.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.10
• /
• pp.1231-1238
• /
• 1999

This paper presents an approach for the optimal heating load Identification using Diagonal Recurrent Neural Networks(DRNN). In this paper, the DRNN captures the dynamic nature of a system and since it is not fully connected, training is much faster than a fully connected recurrent neural network. The architecture of DRNN is a modified model of the fully connected recurrent neural network with one hidden layer. The hidden layer is comprised of self-recurrent neurons, each feeding its output only into itself. In this study, A dynamic backpropagation (DBP) with delta-bar-delta learning method is used to train an optimal heating load identifier. Delta-bar-delta learning method is an empirical method to adapt the learning rate gradually during the training period in order to improve accuracy in a short time. The simulation results based on experimental data show that the proposed model is superior to the other methods in most cases, in regard of not only learning speed but also identification accuracy.

• International Journal of Aeronautical and Space Sciences
• /
• v.18 no.3
• /
• pp.474-484
• /
• 2017

This paper studies the applicability of an efficient numerical model based on artificial neural networks (ANNs) to predict the dynamic responses of the wing structure of an airplane due to atmospheric turbulence in the time domain. The turbulence velocity is given in the form of a stationary Gaussian random process with the von Karman power spectral density. The wing structure is modeled by a classical beam considering bending and torsional deformations. An unsteady vortex-lattice method is applied to estimate the aerodynamic pressure distribution on the wing surface. Initially, the trim condition is obtained, then structural dynamic responses are computed. The numerical solution of the wing structure's responses to a random turbulence profile is used as a training data for the ANN. The current ANN is a three-layer network with the output fed back to the input layer through delays. The results from this study have validated the proposed low-cost ANN model for the predictions of dynamic responses of wing structures due to atmospheric turbulence. The accuracy of the predicted results by the ANN was discussed. The paper indicated that predictions for the bending moments are more accurate than those for the torsional moments of the wing structure.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.7
• /
• pp.675-679
• /
• 2001

In this paper, we present a new competitive learning algorithm called Dynamic Competitive Learning (DCL). DCL is a supervised learning method that dynamically generates output neurons and initializes automatically the weight vectors from training patterns. It introduces a new parameter called LOG (Limit of Grade) to decide whether an output neuron is created or not. If the class of at least one among the LOG number of nearest output neurons is the same as the class of the present training pattern, then DCL adjusts the weight vector associated with the output neuron to learn the pattern. If the classes of all the nearest output neurons are different from the class of the training pattern, a new output neuron is created and the given training pattern is used to initialize the weight vector of the created neuron. The proposed method is significantly different from the previous competitive learning algorithms in the point that the selected neuron for learning is not limited only to the winner and the output neurons are dynamically generated during the learning process. In addition, the proposed algorithm has a small number of parameters, which are easy to be determined and applied to real-world problems. Experimental results for pattern recognition of remote sensing data and handwritten numeral data indicate the superiority of DCL in comparison to the conventional competitive learning methods.

• Journal of Korea Multimedia Society
• /
• v.22 no.1
• /
• pp.10-17
• /
• 2019

In High Dynamic Range (HDR) image processing, tone mapping is the process to compress an input image into a Low Dynamic Range (LDR) image. In most cases, the reason that detail preservation is prior to take over tone mapping is that the dynamic range is significantly different between input and output images. In the case of iCAM06, details are separated by using a bilateral filter, however, it causes noise amplification at the dim surround region. Thus, we suggest that the detail signal, which is separated from the bilateral filter, is combined with the base signal after an adaptive Kalman filter is applied according to the local standard deviation. We confirmed that the proposed method enhances the HDR images quality by checking the noise reduction in a dim surround region.