• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.9
• /
• pp.782-787
• /
• 2001

In this paper, we propose an growing and pruning algorithm to design the adaptive structure of modular wavelet neural network(MWNN) with F-projection and geometric growing criterion. Geometric growing criterion consists of estimated error criterion considering local error and angel criterion which attempts to assign wavelet function that is nearly orthogonal to all other existing wavelet functions. There criteria provide a methodology that a network designer can constructs wavelet neural network according to one s intention. The proposed growing algorithm grows the module and the size of modules. Also, the pruning algorithm eliminates unnecessary node of module or module from constructed MWNN to overcome the problem due to localized characteristics of wavelet neural network which is used to modules of MWNN. We apply the proposed constructing algorithm of the adaptive structure of MWNN to approximation problems of 1-D function and 2-D function, and evaluate the effectiveness of the proposed algorithm.

• Journal of Intelligence and Information Systems
• /
• v.4 no.1
• /
• pp.133-147
• /
• 1998

Recently, many researchers have been involved in finding deterministic equations which can accurately predict future event, based on chaotic theory, or fractal theory. The theory says that some events which seem very random but internally deterministic can be accurately predicted by fractal equations. In contrast to the conventional methods, such as AR model, MA, model, or ARIMA model, the fractal equation attempts to discover a deterministic order inherent in time series data set. In discovering deterministic order, researchers have found that neural networks are much more effective than the conventional statistical models. Even though prediction accuracy of the network can be different depending on the topological structure and modification of the algorithms, many researchers asserted that the neural network systems outperforms other systems, because of non-linear behaviour of the network models, mechanisms of massive parallel processing, generalization capability based on adaptive learning. However, recent survey shows that prediction accuracy of the forecasting models can be determined by the model structure and data structures. In the experiments based on actual economic data sets, it was found that the prediction accuracy of the neural network model is similar to the performance level of the conventional forecasting model. Especially, for the data set which is deterministically chaotic, the AR model, a conventional statistical model, was not significantly different from the MLP model, a neural network model. This result shows that the forecasting model. This result shows that the forecasting model a, pp.opriate to a prediction task should be selected based on characteristics of the time series data set. Analysis of the characteristics of the data set was performed by fractal analysis, measurement of Hurst index, and measurement of Lyapunov exponents. As a conclusion, a significant difference was not found in forecasting future events for the time series data which is deterministically chaotic, between a conventional forecasting model and a typical neural network model.

• Computers and Concrete
• /
• v.13 no.6
• /
• pp.739-748
• /
• 2014

In this article, by using experimental studies and artificial neural network has been tried to investigate the use of nano-silica as concrete admixture to reduce alkali-silica reaction. If there are reactive aggregates and alkali of cement with enough moisture in concrete, a gel will be formed. Then with high reactivity between alkali of cement and existence of silica in aggregates, this gel will expand by absorption of water, and causes expansive pressure and cracks be formed. At the time passes, this gel will reduce both durability and strength of the concrete. By reducing the size of silicate to nano, specific surface area of particles and number of atoms on the surface will be increased, which causes more pozzolanic activity of them. Nano-silica can react with calcium hydroxide ($Ca(OH)_2$) and produces C-S-H gel. In this study, accelerated mortar bar specimens according to ASTM C 1260 and ASTM C 1567, with different mix proportions were prepared using aggregates of Kerman, such as: none admixture and plasticizer, different proportions of nano-silica separately. By opening the moulds after 24 hour and curing in water at $80^{\circ}C$ for 24 hour, then curing in (1N NaOH) at $80^{\circ}C$ for 14 days, length expansion of mortar bars were measured and compared. It was noted that, the lowest length expansion of a specimens shows the best proportion of admixture based on alkali-silica reactivity. Then, prediction of alkali-silica reaction of concrete has been investigated by using artificial neural network. In this study the backpropagation network has been used and compared with different algorithms to train network. Finally, the best amount of nano silica for adding to mix proportion, also the best algorithm and number of neurons in hidden layer of artificial neural network have been offered.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.16 no.2
• /
• pp.179-184
• /
• 2006

An intuitive understanding of the dynamic pattern generation in asymmetric networks may be considered an essential component in developing models for the dynamic information processing. It has been reported that the neural network with cyclic connections generates multiple limit cycles. The dynamics of discrete time network with cyclic connections has been investigated intensively. However, the dynamics of a cyclic connection neural network in continuous-time has not been well-known due to the considerable complexity involved in its calculation. In this paper, the dynamic behavior of a continuous-time cyclic connection neural network, in which each neuron is connected only to its nearest neurons with binary synaptic weights of ${\pm}1$, has been investigated. Furthermore, the dynamics and stability of the network have been analyzed using a piece-wise linear approximation.

• Journal of the Institute of Convergence Signal Processing
• /
• v.4 no.3
• /
• pp.49-54
• /
• 2003

In this paper, we propose a controller with the self-organizing neural network compensator for compensating PID controller's response. PID controller has simple design method but needs a lot of trials and errors to determine coefficients. A neural network control method does not have optimal structure as the parameters are pre-specified by designers. In this paper, to solve this problem, we use a self-organizing neural network which has Back Propagation Network algorithm using a Gaussian Potential Function as an activation function of hidden layer nodes for compensating PID controller's output. Self-Organizing Neural Network's learning is proceeded by Gaussian Function's Mean, Variance and number which are automatically adjusted. As the results of simulation through the second order plant, we confirmed that the proposed controller get a good response compare with a PID controller. And we implemented the of controller performance hydraulic servo motor system using the DSP processor. Then we observed an experimental results.

• Proceedings of the KAIS Fall Conference
• /
• 2008.05a
• /
• pp.248-250
• /
• 2008

This paper proposes a prediction method to prevent traffic accident and reduce to vehicle waiting time using neural network. Computer simulation results proved reducing average vehicle waiting time which proposed coordinating green time better than electro-sensitive traffic light system dose not consider coordinating green time. Moreover, we present neural network approach for traffic accident prediction with unnormalized (actual or original collected) data. This approach is not consider the maximum value of data and possible use the network without normalizing but the predictive accuracy is better. Also, the unnormalized method shows better predictive accuracy than the normalized method given by maximum value. Therefore, we can make the best use of this model in software reliability prediction using unnormalized data. Computer simulation results proved reducing traffic accident waiting time which proposed neural network better than conventional system dosen't consider neural network.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.12
• /
• pp.1588-1594
• /
• 2020

In this paper, we discuss a data balancing method for learning a neural network that generates digital holograms using a deep neural network (DNN). Deep neural networks are based on deep learning (DL) technology and use a generative adversarial network (GAN) series. The fringe pattern, which is the basic unit of a hologram to be created through a deep neural network, has very different data types depending on the hologram plane and the position of the object. However, because the criteria for classifying the data are not clear, an imbalance in the training data may occur. The imbalance of learning data acts as a factor of instability in learning. Therefore, it presents a method for classifying and balancing data for which the classification criteria are not clear. And it shows that learning is stabilized through this.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.3
• /
• pp.51-57
• /
• 2021

In this paper, we propose a method to find the DoF(Depth of field) that is blurred in an image by focusing and out-focusing the camera through a efficient convolutional neural network. Our approach uses the RGB channel-based cross-correlation filter to efficiently classify the DoF region from the image and build data for learning in the convolutional neural network. A data pair of the training data is established between the image and the DoF weighted map. Data used for learning uses DoF weight maps extracted by cross-correlation filters, and uses the result of applying the smoothing process to increase the convergence rate in the network learning stage. The DoF weighted image obtained as the test result stably finds the DoF region in the input image. As a result, the proposed method can be used in various places such as NPR(Non-photorealistic rendering) rendering and object detection by using the DoF area as the user's ROI(Region of interest).

• International Journal of Computer Science & Network Security
• /
• v.24 no.5
• /
• pp.11-20
• /
• 2024

While artificial neural networks are adept at identifying patterns, they can struggle to distinguish between actual correlations and false associations between extracted facial features and criminal behavior within the training data. These associations may not indicate causal connections. Socioeconomic factors, ethnicity, or even chance occurrences in the data can influence both facial features and criminal activity. Consequently, the artificial neural network might identify linked features without understanding the underlying cause. This raises concerns about incorrect linkages and potential misclassification of individuals based on features unrelated to criminal tendencies. To address this challenge, we propose a novel region-based training approach for artificial neural networks focused on criminal propensity detection. Instead of solely relying on overall facial recognition, the network would systematically analyze each facial feature in isolation. This fine-grained approach would enable the network to identify which specific features hold the strongest correlations with criminal activity within the training data. By focusing on these key features, the network can be optimized for more accurate and reliable criminal propensity prediction. This study examines the effectiveness of various algorithms for criminal propensity classification. We evaluate YOLO versions YOLOv5 and YOLOv8 alongside VGG-16. Our findings indicate that YOLO achieved the highest accuracy 0.93 in classifying criminal and non-criminal facial features. While these results are promising, we acknowledge the need for further research on bias and misclassification in criminal justice applications

• Journal of Broadcast Engineering
• /
• v.24 no.4
• /
• pp.633-642
• /
• 2019

Recently, the convolution neural network (CNN) model at a single image super-resolution (SISR) have been very successful. The residual learning method can improve training stability and network performance in CNN. In this paper, we propose a SISR using recursive residual network architecture by introducing dense skip connections for learning nonlinear mapping from low-resolution input image to high-resolution target image. The proposed SISR method adopts a method of the recursive residual learning to mitigate the difficulty of the deep network training and remove unnecessary modules for easier to optimize in CNN layers because of the concise and compact recursive network via dense skip connection method. The proposed method not only alleviates the vanishing-gradient problem of a very deep network, but also get the outstanding performance with low complexity of neural network, which allows the neural network to perform training, thereby exhibiting improved performance of SISR method.