• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.4
• /
• pp.1008-1018
• /
• 1996

For assigning channel in a cellular sireless network, we purpose satisfying constraints with a related probability between cells and channels on the channel interference matrix formed by m terminals and n channels. And we purpose to get efficient channel assigning to limited channels in a wireless cell using parallelism of neural networks. In this paper, we solve the problem according to the number of requeirements when channel change with 11-533 procession elements. We demonstrate efficiency of proposed algorithm through same simulations in a specific time period.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2003.06a
• /
• pp.145-148
• /
• 2003

In this paper an nonlinear neuro PID controller is constructed by the control system of general PID controller using a Self-Recurrent Neural Network. And the games of the PID controller in the proposed control system are automatically adjusted by back-propagation algorithm of the neural network. Applying to the position control system, it's performance is verified through the results of computer simulation.

• Journal of KIISE:Software and Applications
• /
• v.27 no.11
• /
• pp.1082-1087
• /
• 2000

신경회로망 구조의 정제(精製)는 회로망의 일반화능력이나 효율성의 관점에서 중요한 문제이다. 본 논문에서는 feed-forward neural networks로부터 은닉지식을 추출하는 방법을 사용하여 네트워크 재구성을 통한 정제방법을 제안한다. 먼저, 효율적인 if-then rule 추출방법을 제시하고 그 추출된 룰들을 사용하여 룰기반 네트워크로 변환하는 과정을 보여준다. 생성된 룰기반 네트워크 fully connected network에 비하여 상당히 축소된 연결 복잡도를 가지게 되며 일반적으로 더 우수한 일반화능력을 가지게 된다. 본 연구는 도메인 지식이 없이 데이타만 사용하여 어떻게 정제된 룰기반 신경망회로를 생성하고 있는가를 보여준다. 도메인 데이타들에 대한 실험결과도 제시하였다.

• Annual Conference on Human and Language Technology
• /
• 1989.10a
• /
• pp.267-271
• /
• 1989

최근 Scofield는 coulomb energy network에 적용할 수 있는 learning algorithm(supervised learning algorithm)을 제안하였다. 이 learning algorithm은 multi-layer network에도 쉽게 적용이 가능하고 한 layer 에서 발생한 error가 다른 layer에 영향을 주지 않아서 system을 modular하게 구성할 수가 있으며 각 layer를 독립적으로 learning 시킬 수 있는 특징이 있다. 본 논문에서는 coulomb energy network를 이용하여 한글인식을 위한 neural network를 구현하여 인식실험을 한 결과와 구현한 network 에서 인식율을 높이기 위한 방안 (2 stage learning) 을 제시한다.

• Structural Engineering and Mechanics
• /
• v.16 no.1
• /
• pp.83-99
• /
• 2003

The aim of this paper is to develop the ISCOSTFUN (Intelligent System for Prediction of Concrete Strength by Functional Networks) in order to provide in-place strength information of the concrete to facilitate concrete from removal and scheduling for construction. For this purpose, the system is developed using Functional Network (FN) by learning functions instead of weights as in Artificial Neural Networks (ANN). In serial functional network, the functions are trained from enough input-output data and the input for one functional network is the output of the other functional network. Using ISCOSTFUN it is possible to predict early strength as well as 7-day and 28-day strength of concrete. Altogether seven functional networks are used for prediction of strength development. This study shows that ISCOSTFUN using functional network is very efficient for predicting the compressive strength development of concrete and it takes less computer time as compared to well known Back Propagation Neural Network (BPN).

• Journal of Institute of Control, Robotics and Systems
• /
• v.8 no.2
• /
• pp.120-125
• /
• 2002

An optimization approach is used to partition the field of view. A cost function is defined to represent the constraints on the solution, which is then mapped onto a two-dimensional Hopfield neural network for minimization. Each neuron in the network represents a possible match between a field of view and one or multiple objects. Partition is achieved by initializing each neuron that represents a possible match and then allowing the network to settle down into a stable state. The network uses the initial inputs and the compatibility measures between a field of view and one or multiple objects to find a stable state.

• Korean Journal of Remote Sensing
• /
• v.36 no.5_2
• /
• pp.961-973
• /
• 2020

In this manuscript, a new pansharpening model based on Convolutional Neural Network (CNN) was developed. Dilated convolution, which is one of the representative convolution technologies in CNN, was applied to the model by making it deep and complex to improve the performance of the deep learning architecture. Based on the dilated convolution, the residual network is used to enhance the efficiency of training process. In addition, we consider the spatial correlation coefficient in the loss function with traditional L1 norm. We experimented with Dilated Residual Networks (DRNet), which is applied to the structure using only a panchromatic (PAN) image and using both a PAN and multispectral (MS) image. In the experiments using KOMPSAT-3A, DRNet using both a PAN and MS image tended to overfit the spectral characteristics, and DRNet using only a PAN image showed a spatial resolution improvement over existing CNN-based models.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.10 no.2
• /
• pp.15-22
• /
• 2006

An intelligent performance diagnostic program using the Neural Network was proposed for PW206C turboshaft engine. It was selected as a power plant for the tilt rotor type Smart UAV(Unmanned Aerial Vehicle) which is being developed by KARI (Korea Aerospace Research Institute). For teeming the NN(Neural Network), a BPN(Back Propagation Network) with one hidden, one input and one output layer was used. The input layer has seven neurons: variations of measurement parameters such as SHP, MF, P2, T2, P4, T4 and T5, and the output layer uses 6 neurons: degradation ratios of flow capacities and efficiencies for compressor, compressor turbine and power turbine, respectively, Database for network teaming and test was constructed using a gas turbine performance simulation program. From application of the learned networks to diagnostics of the PW206C turboshaft engine, it was confirmed that the proposed diagnostics algorithm could detect well the single fault types such as compressor fouling and compressor turbine erosion.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.12
• /
• pp.1772-1781
• /
• 2017

In this study, the design methodology for alleviating the overfitting problem of Polynomial Neural Networks(PNN) is realized with the aid of two kinds techniques such as L2 regularization and Sum of Squared Coefficients (SSC). The PNN is widely used as a kind of mathematical modeling methods such as the identification of linear system by input/output data and the regression analysis modeling method for prediction problem. PNN is an algorithm that obtains preferred network structure by generating consecutive layers as well as nodes by using a multivariate polynomial subexpression. It has much fewer nodes and more flexible adaptability than existing neural network algorithms. However, such algorithms lead to overfitting problems due to noise sensitivity as well as excessive trainning while generation of successive network layers. To alleviate such overfitting problem and also effectively design its ensuing deep network structure, two techniques are introduced. That is we use the two techniques of both SSC(Sum of Squared Coefficients) and $L_2$ regularization for consecutive generation of each layer's nodes as well as each layer in order to construct the deep PNN structure. The technique of $L_2$ regularization is used for the minimum coefficient estimation by adding penalty term to cost function. $L_2$ regularization is a kind of representative methods of reducing the influence of noise by flattening the solution space and also lessening coefficient size. The technique for the SSC is implemented for the minimization of Sum of Squared Coefficients of polynomial instead of using the square of errors. In the sequel, the overfitting problem of the deep PNN structure is stabilized by the proposed method. This study leads to the possibility of deep network structure design as well as big data processing and also the superiority of the network performance through experiments is shown.

• Journal of KIISE:Software and Applications
• /
• v.34 no.7
• /
• pp.585-594
• /
• 2007

The back propagation algorithm took a long time to learn the input patterns and was difficult to train the additional or repeated learning patterns. So Aleksander proposed the binary neural network which could overcome the disadvantages of BP Network. But it had the limitation of repeated learning and was impossible to extract a generalized pattern. In this paper, we proposed a dynamic 3 dimensional Neuro System which was consisted of a learning network which was based on weightless neural network and a feedback module which could accumulate the characteristic. The proposed system was enable to train additional and repeated patterns. Also it could be produced a generalized pattern by putting a proper threshold into each learning-net's discriminator which was resulted from learning procedures. And then we reused the generalized pattern to elevate the recognition rate. In the last processing step to decide right category, we used maximum response detector. We experimented using the MNIST database of NIST and got 99.3% of right recognition rate for training data.