• International Journal of Computer Science & Network Security
• /
• v.23 no.1
• /
• pp.41-46
• /
• 2023

Because of the significant role that harakat plays in Arabic text, this paper used deep learning to extract Arabic text with its harakat from an image. Convolutional neural networks and recurrent neural network algorithms were applied to the dataset, which contained 110 images, each representing one word. The results showed the ability to extract some letters with harakat.

• International journal of advanced smart convergence
• /
• v.8 no.1
• /
• pp.141-146
• /
• 2019

These days, there are many applications using neural networks as parts of their system. On the other hand, adversarial examples have become an important issue concerining the security of neural networks. A classifier in neural networks can be fooled and make it miss-classified by adversarial examples. There are many research to encounter adversarial examples by using denoising methods. Some of them using GAN (Generative Adversarial Network) in order to remove adversarial noise from input images. By producing an image from generator network that is close enough to the original clean image, the adversarial examples effects can be reduced. However, there is a chance when adversarial noise can survive the approximation process because it is not like a normal noise. In this chance, we propose a research that utilizes high-level representation in the classifier by combining GAN network with a trained U-Net network. This approach focuses on minimizing the loss function on high representation terms, in order to minimize the difference between the high representation level of the clean data and the approximated output of the noisy data in the training dataset. Furthermore, the generated output is checked whether it shows minimum error compared to true label or not. U-Net network is trained with true label to make sure the generated output gives minimum error in the end. At last, the remaining adversarial noise that still exist after low-level approximation can be removed with the U-Net, because of the minimization on high representation terms.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.5 no.1
• /
• pp.52-58
• /
• 2005

We have developed a Group Method of Data Handling (GMDH) type algorithm for designing multi-layered neural networks. The algorithm is general enough that it will accept any number of inputs and any sized training set. Each neuron of the resulting network is a function of two of the inputs to the layer. The equation for each of the neurons is a quadratic polynomial. Several forms of the equation are tested for each neuron to make sure that only the best equation of two inputs is kept. All possible combinations of two inputs to each layer are also tested. By carefully testing each resulting neuron, we have developed an algorithm to keep only the best neurons at each level. The algorithm's goal is to create as accurate a network as possible while minimizing the size of the network. Software was developed to train and simulate networks using our algorithm. Several applications were modeled using our software, and the result was that our algorithm succeeded in developing small, accurate, multi-layer networks.

• Proceedings of the Korean Information Science Society Conference
• /
• 1999.10b
• /
• pp.327-332
• /
• 1999

Fuzzy systems, neural networks and genetic algorithms have different origins and thus have differently developed their own unique characteristics. These characteristics can be used as a good complement to the others. Therefore, many researches have been devoted to not only these techniques but also fusion of them. This paper briefly summarizes these three techniques and surveys the researches on fusion of them.

• Proceedings of the Acoustical Society of Korea Conference
• /
• autumn
• /
• pp.527-530
• /
• 2004

We present a recursive total least squares (RTLS) algorithm for multilayer feedforward neural networks. So far, recursive least squares (RLS) has been successfully applied to training multilayer feedforward neural networks. But, when input data contain additive noise, the results from RLS could be biased. Such biased results can be avoided by using the recursive total least squares (RTLS) algorithm. The RTLS algorithm described in this paper gives better performance than RLS algorithm over a wide range of SNRs and involves approximately the same computational complexity of $O(N^{2})$.

• Journal of Power Electronics
• /
• v.11 no.6
• /
• pp.922-930
• /
• 2011

In this paper, two radial basis function neural networks (RBFNNs) are used to dynamically identify harmonics content in converter waveforms based on the p-q (real power-imaginary power) theory. The converter waveforms are analyzed and the types of harmonic content are identified over a wide operating range. Constant power and sinusoidal current compensation strategies are investigated in this paper. The RBFNN filtering training algorithm is based on a systematic and computationally efficient training method called the hybrid learning method. In this new methodology, the RBFNN is combined with the p-q theory to extract the harmonics content in converter waveforms. The small size and the robustness of the resulting network models reflect the effectiveness of the algorithm. The analysis is verified using MATLAB simulations.

• Journal of Broadcast Engineering
• /
• v.26 no.7
• /
• pp.868-876
• /
• 2021

Deploying deep neural network model training performs remarkable performance in the fields of Object detection and Instance segmentation. To train these models, features are first extracted from the input image using a backbone network. The extracted features can be reused by various tasks. Research has been actively conducted to serve various tasks by using these learned features. In this process, standardization discussions about encoding, decoding, and transmission methods are proceeding actively. In this scenario, it is necessary to analyze the response characteristics of features against various distortions that may occur in the data transmission or data compression process. In this paper, experiment was conducted to inject various distortions into the feature in the object recognition task. And analyze the mAP (mean Average Precision) metric between the predicted value output from the neural network and the target value as the intensity of various distortions was increased. Experiments have shown that features are more robust to distortion than images. And this points out that using the feature as transmission means can prevent the loss of information against the various distortions during data transmission and compression process.

• Journal of the Semiconductor & Display Technology
• /
• v.22 no.2
• /
• pp.128-132
• /
• 2023

Recently, with the development of deep learning technology, research on pedestrian attribute recognition technology using deep neural networks has been actively conducted. Existing pedestrian attribute recognition techniques can be obtained in such a way as global-based, regional-area-based, visual attention-based, sequential prediction-based, and newly designed loss function-based, depending on how pedestrian attributes are detected. It is known that the performance of these pedestrian attribute recognition technologies varies greatly depending on the type of backbone network that constitutes the deep neural networks model. Therefore, in this paper, several backbone networks are applied to the baseline pedestrian attribute recognition model and the performance changes of the model are analyzed. In this paper, the analysis is conducted using Resnet34, Resnet50, Resnet101, Swin-tiny, and Swinv2-tiny, which are representative backbone networks used in the fields of image classification, object detection, etc. Furthermore, this paper analyzes the change in time complexity when inferencing each backbone network using a CPU and a GPU.

• Journal of the Korean Society of Industry Convergence
• /
• v.23 no.3
• /
• pp.515-521
• /
• 2020

Deep learning methods have been effectively used to provide great improvement in various research fields such as machine learning, image processing and computer vision. One of the most frequently used deep learning methods in image processing is the convolutional neural networks. Compared to the traditional artificial neural networks, convolutional neural networks do not use the predefined kernels, but instead they learn data specific kernels. This property makes them to be used as feature extractors as well. In this study, we compared the quality of CNN features for traditional texture feature extraction methods. Experimental results demonstrate the superiority of the CNN features. Additionally, the recognition process and result of a pioneering CNN on MNIST database are presented.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.43 no.6
• /
• pp.991-1000
• /
• 1994

To control nonlinear systems adaptively, we improve learning speed of neural networks and present a novel control algorithm characterized by compensation of control inputs. In an error-backpropagation algorithm for tranining multilayer neural networks(MLNN's) the effect of the slope of activation functions on learning performance is investigated and the learning speed of neural networks is improved by auto-adjusting the slope of activation functions. The control system is composed of two MLNN's, one for control and the other for identification, with the weights initialized by off-line training. The control algoritm is modified by a control strategy which compensates the control error induced by the indentification error. Computer simulations show that the proposed control algorithm is efficient in controlling a nonlinear system with abruptly changing parameters.