• Journal of Industrial Technology
• /
• v.41 no.1
• /
• pp.1-6
• /
• 2021

Fly ash and silica fume belong to industry by-products that can be used to produce concrete. This study shows the model of a neural network to evaluate the strength development of blended concrete containing fly ash and silica fume. The neural network model has four input parameters, such as fly ash replacement content, silica fume replacement content, water/binder ratio, and ages. Strength is the output variable of neural network. Based on the backpropagation algorithm, the values of elements in the hidden layer of neural network are determined. The number of neurons in the hidden layer is confirmed based on trial calculations. We find (1) neural network can give a reasonable evaluation of the strength development of composite concrete. Neural network can reflect the improvement of strength due to silica fume additions and can consider the reductions of strength as water/binder increases. (2) When the number of neurons in the hidden layer is five, the prediction results show more accuracy than four neurons in the hidden layer. Moreover, five neurons in the hidden layer can reproduce the strength crossover between fly ash concrete and plain concrete. Summarily, the neural network-based model is valuable for design sustainable composite concrete containing silica fume and fly ash.

• The Journal of Information Systems
• /
• v.9 no.1
• /
• pp.27-44
• /
• 2000

The structure of neural networks is represented by a weighted directed graph with nodes representing units and links representing connections. Each link is assigned a numerical value representing the weight of the connection. In learning process, the values of weights are adjusted by errors. Following experiment results, the interval of adjusting weights, that is, epoch size influenced neural networks＇ performance. As epoch size is larger than a certain size, neural networks＇performance decreased drastically. And the number of hidden layer＇s node also influenced neural networks＇performance. The networks＇performance decreased as hidden layers have more nodes and then increased at some number of hidden layer＇s node. So, in implementing of neural networks the epoch size and the number of hidden layer＇s node should be decided by systematic methods, not empirical or heuristic methods.

• Journal of KIISE:Software and Applications
• /
• v.28 no.9
• /
• pp.603-611
• /
• 2001

This paper proposes a novel fast layer-by-layer algorithm that has better generalization capability. In the proposed algorithm, the weights of the hidden layer are updated by the target vector of the hidden layer obtained by least squares method. The proposed algorithm improves the learning speed that can occur due to the small magnitude of the gradient vector in the hidden layer. This algorithm was tested in a handwritten digits recognition problem. The learning speed of the proposed algorithm was faster than those of error back propagation algorithm and modified error function algorithm, and similar to those of Ooyen's method and layer-by-layer algorithm. Moreover, the simulation results showed that the proposed algorithm had the best generalization capability among them regardless of the number of hidden nodes. The proposed algorithm has the advantages of the learning speed of layer-by-layer algorithm and the generalization capability of error back propagation algorithm and modified error function algorithm.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.28B no.11
• /
• pp.852-858
• /
• 1991

An adaptive hierarchical classifier(AHCL) for Korean character recognition using a neural net is designed. This classifier has two neural nets: USACL (Unsupervised Adaptive Classifier) and SACL (Supervised Adaptive Classifier). USACL has the input layer and the output layer. The input layer and the output layer are fully connected. The nodes in the output layer are generated by the unsupervised and nearest neighbor learning rule during learning. SACL has the input layer, the hidden layer and the output layer. The input layer and the hidden layer arefully connected, and the hidden layer and the output layer are partially connected. The nodes in the SACL are generated by the supervised and nearest neighbor learning rule during learning. USACL has pre-attentive effect, which perform partial search instead of full search during SACL classification to enhance processing speed. The input of USACL and SACL is a directional edge feature with a directional receptive field. In order to test the performance of the AHCL, various multi-font printed Hangul characters are used in learning and testing, and its processing its speed and and classification rate are compared with the conventional LVQ(Learning Vector Quantizer) which has the nearest neighbor learning rule.

• Journal of Korea Multimedia Society
• /
• v.6 no.7
• /
• pp.1312-1320
• /
• 2003

In this paper, we consider a multilayer neural network, with a single hidden layer. Error backpropagation learning method used widely in multilayer neural networks has a possibility of local minima due to the inadequate weights and the insufficient number of hidden nodes. So we propose a fuzzy supervised learning algorithm by using self-generation that self-generates hidden nodes by the compound fuzzy single layer perceptron and modified ART1. From the input layer to hidden layer, a modified ART1 is used to produce nodes. And winner take-all method is adopted to the connection weight adaptation, so that a stored pattern for some pattern gets updated. The proposed method has applied to the student identification card images. In simulation results, the proposed method reduces a possibility of local minima and improves learning speed and paralysis than the conventional error backpropagation learning algorithm.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.9
• /
• pp.1137-1153
• /
• 1995

It is not yet possible to solve the optimal number of neurons in hidden layer at neural networks. However, it has been proposed and proved by experiments that there is a limit in increasing the number of neuron in hidden layer, because too much incrememt will cause instability,local minima and large error. This paper proposes a module neural controller with pattern recognition ability to solve the above trade-off problems and to obtain fast learning convergence speed. The proposed neural controller is composed of several module having Multi-layer Perrceptron(MLP). Each module have the less neurons in hidden layer, because it learns only input patterns having a similar learning directions. Experiments with six joint robot manipulator have shown the effectiveness and the feasibility of the proposed the parallel module neural controller with pattern recognition perceptron.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.3B
• /
• pp.565-569
• /
• 2000

We compose the two-stage recurrent neural network that returns both signals of a hidden and an output layer to the hidden layer. It is tested on the basis of syllables for Korean spoken digit from /gong/to /gu. For these experiments, we adjust the neuron number of the hidden layer, the predictive order of input data and self-recurrent coefficient of the decision state layer. By the experimental results, the recognition rate of this neural network is between 91% and 97.5% in the speaker-dependent case and between 80.75% and 92% in the speaker-independent case. In the speaker-dependent case, this network shows an equivalent recognition performance to Jordan and Elman network but in the speaker-independent case, it does improved performance.

• International Journal of Control, Automation, and Systems
• /
• v.3 no.2
• /
• pp.202-210
• /
• 2005

This paper presents a new neural network control for nonlinear bridge systems with earthquake excitation. We design multi-layer neural network controllers with a single hidden layer. The selection of an optimal number of neurons in the hidden layer is an important design step for control performance. To select an optimal number of hidden neurons, we progressively add one hidden neuron and observe the change in a performance measure given by the weighted sum of the system error and the control force. The number of hidden neurons which minimizes the performance measure is selected for implementation. A neural network was trained for mitigating vibrations of bridge systems caused by El Centro earthquake. We applied the proposed control approach to a single-degree-of-freedom (SDOF) and a two-degree-of-freedom (TDOF) bridge system. We assessed the robustness of the control system using randomly generated earthquake excitations which were not used in training the neural network. Our results show that the neural network controller drastically mitigates the effect of the disturbance.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.35C no.3
• /
• pp.79-89
• /
• 1998

The Error Back-Propagation(EBP) algorithm is widely applied to train a multi-layer perceptron, which is a neural network model frequently used to solve complex problems such as pattern recognition, adaptive control, and global optimization. However, the EBP is basically a gradient descent method, which may get stuck in a local minimum, leading to failure in finding the globally optimal solution. Moreover, a multi-layer perceptron suffers from locking a systematic determination of the network structure appropriate for a given problem. It is usually the case to determine the number of hidden nodes by trial and error. In this paper, we propose a new algorithm to efficiently train a multi-layer perceptron. OUr algorithm uses stochastic perturbation in the weight space to effectively escape from local minima in multi-layer perceptron learning. Stochastic perturbation probabilistically re-initializes weights associated with hidden nodes to escape a local minimum if the probabilistically re-initializes weights associated with hidden nodes to escape a local minimum if the EGP learning gets stuck to it. Addition of new hidden nodes also can be viewed asa special case of stochastic perturbation. Using stochastic perturbation we can solve the local minima problem and the network structure design in a unified way. The results of our experiments with several benchmark test problems including theparity problem, the two-spirals problem, andthe credit-screening data show that our algorithm is very efficient.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.130-138
• /
• 2021

This paper presents an effective structure by applying various normalization to Convolutional Neural Networks (CNN) for seismic event classification. Normalization techniques can not only improve the learning speed of neural networks, but also show robustness to noise. In this paper, we analyze the effect of input data normalization and hidden layer normalization on the deep learning model for seismic event classification. In addition an effective model is derived through various experiments according to the structure of the applied hidden layer. As a result of various experiments, the model that applied input data normalization and weight normalization to the first hidden layer showed the most stable performance improvement.