• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.43.2-43
• /
• 2001

In this paper, we consider the problems of the evolutionary designed neural networks structures by genetic algorithm. Neural networks has been applied to various application fields since back-propagation algorithm was proposed, e.g. function approximation, pattern or character recognition and so on. However, one of difficulties to use the neural networks. It is how to design the structure of the neural network. Researchers and users design networks structures and training parameters such as learning rate and momentum rate and so on, by trial and error based on their experiences. In the case of designing large scales neural networks, it is very hard work for manually design by try and error. For this difficulty, various structural learning algorithms have been proposed. Especially, the technique of using genetic algorithm for networks structures design has been ...

• Korean Journal of Remote Sensing
• /
• v.26 no.6
• /
• pp.693-703
• /
• 2010

High spectral resolution of hyperspectral data enables analysis of complex natural phenomena that is reflected on the data nonlinearly. Although many manifold learning methods have been developed for such problems, most methods do not consider the spatial correlation between samples that is inherent and useful in remote sensing data. We propose a manifold learning method which directly combines the spatial proximity and the spectral similarity through kernel PCA framework. A gain factor caused by spatial proximity is first modelled with a heat kernel, and is added to the original similarity computed from the spectral values of a pair of samples. Parameters are tuned with intelligent grid search (IGS) method for the derived manifold coordinates to achieve optimal classification accuracies. Of particular interest is its performance with small training size, because labelled samples are usually scarce due to its high acquisition cost. The proposed spatial kernel PCA (KPCA) is compared with PCA in terms of classification accuracy with the nearest-neighbourhood classification method.

• Geomechanics and Engineering
• /
• v.1 no.4
• /
• pp.307-321
• /
• 2009

The paper employs a feed forward neural network with back-propagation algorithm for modeling time dependent swell in clays containing carbonate in the presence of sulfuric acid. The oedometer swell percent is estimated at a nominal surcharge pressure of 6.25 kPa to develop 612 data sets for modeling. The input parameters used in the network include time, sulfuric acid concentration, carbonate percentage, and liquid limit. Among the total data sets, 280 (46%) were assigned to training, 175 (29%) for testing and the remaining 157 data sets (25%) were relegated to cross validation. The network was programmed to process this information and predict the percent swell at any time, knowing the variable involved. The study demonstrates that it is possible to develop a general BPNN model that can predict time dependent swell with relatively high accuracy with observed data ($R^2$=0.9986). The obtained results are also compared with generated non-linear regression model.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.790-809
• /
• 2019

The design of an efficient scaler to enhance the edge data is one of the most important issues in video signal applications, because the perceptual quality of the processed image is sensitively affected by the degradation of edge data. Various conventional scaling schemes have been proposed to enhance the edge data. In this paper, we propose an efficient scaling algorithm for this purpose. The proposed method is based on the discrete cosine transform-based interpolation filter (DCT-IF) because it outperforms other scaling algorithms in various configurations. The proposed DCT-IF incorporates weighting parameters that are optimized for training data. Simulation results show that the quality of the resized image produced by the proposed DCT-IF is much higher than that of those produced by the conventional schemes, although the proposed DCT-IF is more complex than other conventional scaling algorithms.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.98-101
• /
• 2019

At present, deep convolutional neural networks have made a very important contribution in single-image super-resolution. Through the learning of the neural networks, the features of input images are transformed and combined to establish a nonlinear mapping of low-resolution images to high-resolution images. Some previous methods are difficult to train and take up a lot of memory. In this paper, we proposed a simple and compact deep recursive residual network learning the features for single image super resolution. Global residual learning and local residual learning are used to reduce the problems of training deep neural networks. And the recursive structure controls the number of parameters to save memory. Experimental results show that the proposed method improved image qualities that occur in previous methods.

• Journal of information and communication convergence engineering
• /
• v.16 no.4
• /
• pp.264-270
• /
• 2018

In our research study, we aim at optimizing multiple load in cloud, effective resource allocation and lesser response time for the job assigned. Using Hadoop on datacenter is the best and most efficient analytical service for any corporates. To provide effective and reliable performance analytical computing interface to the client, various cloud service providers host Hadoop clusters. The previous works done by many scholars were aimed at execution of workflows on Hadoop platform which also minimizes the cost of virtual machines and other computing resources. Earlier stochastic hill climbing technique was applied for single parameter and now we are working to optimize multiple parameters in the cloud data centers with proposed heuristic hill climbing. As many users try to priorities their job simultaneously in the cluster, resource optimized workflow scheduling technique should be very reliable to complete the task assigned before the deadlines and also to optimize the usage of the resources in cloud.

• Nuclear Engineering and Technology
• /
• v.53 no.11
• /
• pp.3764-3771
• /
• 2021

Based on the convolutional neural network (CNN), a novel technique is investigated for lost gamma source detection in a room. The CNN is trained with the result of a GEANT4 simulation containing a gamma source inside a meshed room. The dataset for the training process is the deposited energy in the meshes of different n-step paths. The neural network is optimized with parameters such as the number of input data and path length. Based on the proposed method, the place of the gamma source can be recognized with reasonable accuracy without human intervention. The results show that only by 5 measurements of the energy deposited in a 5-step path, (5 sequential points 50 cm apart within 1600 meshes), the gamma source location can be estimated with 94% accuracy. Also, the method is tested for the room geometry containing the interior walls. The results show 90% accuracy with the energy deposition measurement in the meshes of a 5-step path.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.9 no.5
• /
• pp.463-471
• /
• 1999

This paper presents a vector controlled induction motor is implemented by neural networks system compared with PI controller for the speed control. The design employed the training strategy with Neural Network Controller(NNC) and Neural Network Emulator(NNE) for speed. In order to update the weights of the controller First of all Emulator updates its parameters by identifying the motor input and output next it supplies the error path to the output stage of the controller using backpropagation algorithm, As Controller produces an adequate output to the system due to neural networks learning capability Vector controlled induction motor characteristics actual motor speed with based on neural network system follows the reference speed better than that of linear PI speed controller.

• Journal of the Semiconductor & Display Technology
• /
• v.19 no.4
• /
• pp.71-76
• /
• 2020

In this paper, we propose a super-resolution method that reconstructs compressed low-resolution images into high-resolution images. We propose a CNN model with a small number of parameters, and even if quantization is applied to the proposed model, super-resolution can be implemented without deteriorating the image quality. To further improve the quality of the compressed low-resolution image, a new degradation model was proposed instead of the existing bicubic degradation model. The proposed degradation model is used only in the training process and can be applied by changing only the parameter values to the original CNN model. In the super-resolution image applying the proposed degradation model, visual artifacts caused by image compression were effectively removed. As a result, our proposed method generates higher PSNR values at compressed images and shows better visual quality, compared to conventional CNN-based SR methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.2
• /
• pp.327-329
• /
• 2021

Underwater acoustic communication channel is influenced by environmental parameters such as multipath, background noise and scattering. Therefore, a transmitted signal is influenced by the sea surface and the sea bottom boundaries, and a received signal shows a delay spread. These factors create a noise in the image and degrade the quality of underwater acoustic communication. To solve these problems, in this paper, we evaluate the performance of an underwater acoustic communication model using a denoising auto-encoder used for unsupervised learning. Noise images generated by the underwater multipath channel were collected and used as training data. Experimental results were analyzed as a PSNR parameter that expressed the noise ratio of the two images.