• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.10
• /
• pp.25-34
• /
• 1999

In this paper, an algorithm is presented to recognize lane and obstacles based on highway road image. The road images obtained by a video camera undergoes a pre-processing that includes filtering, edge detection, and identification of lanes. After this pre-processing, a part of image is grouped into 27 sub-windows and fed into a three-layer feed-forward neural network. The neural network is trained to indicate the road direction and the presence of absence of an obstacle. The proposed algorithm has been tested with the images different from the training images, and demonstrated its efficacy for recognizing lane and obstacles. Based on the test results, it can be said that the algorithm successfully combines the traditional image processing and the neural network principles towards a simpler and more efficient driver warning of assistance system

• Proceedings of the Korean Nuclear Society Conference
• /
• 1997.10a
• /
• pp.58-63
• /
• 1997

This paper is concerned with an algorithm based on neural networks to calculate the axial power distribution using excore defector signals in the nuclear reactor core. The fundamental basis of the algorithm is that the detector response can be fairly accurately estimated using computational codes. In other words, the training set, which represents relationship between detector signals and axial power distributions, for the neural network can be obtained through calculations instead of measurements. Application of the new method to the Yonggwang nuclear power plant unit 3 (YGN-3) shows that it is superior to the current algorithm in place.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.5 no.3
• /
• pp.83-91
• /
• 1995

A neural network with one or more layers of hidden units can be trained using the well-known error back propagation algorithm. According to this algorithm, the synaptic weights of the network are updated during the training by propagating back the error between the expected output and the output provided by the network. However, the error back propagation algorithm is characterized by slow convergence and the time required for training and, in some situation, can be trapped in local minima. A theoretical formulation of a new fast learning method based on tabu search method is presented in this paper. In contrast to the conventional back propagation algorithm which is based solely on the modification of connecting weights of the network by trial and error, the present method involves the calculation of the optimum weights of neural network. The effectiveness and versatility of the present method are verified by the XOR problem. The present method excels in accuracy compared to that of the conventional method of fixed values.

• The Korean Journal of Applied Statistics
• /
• v.26 no.5
• /
• pp.847-856
• /
• 2013

We propose a modified learning process for generalized neural network using a learning algorithm by Liu et al. (2001). We consider the effect of initial weights, training results and learning errors using a modified learning process. We employ an incremental training procedure where training patterns are learned systematically. Our algorithm starts with a single training pattern and a single hidden layer neuron. During the course of neural network training, we try to escape from the local minimum by using a weight scaling technique. We allow the network to grow by adding a hidden layer neuron only after several consecutive failed attempts to escape from a local minimum. Our optimization procedure tends to make the network reach the error tolerance with no or little training after the addition of a hidden layer neuron. Simulation results with suitable initial weights indicate that the present constructive algorithm can obtain neural networks very close to minimal structures and that convergence to a solution in neural network training can be guaranteed. We tested these algorithms extensively with small training sets.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.12
• /
• pp.108-115
• /
• 1994

Most pattern classifiers have been designed based on the ML (Maximum Likelihood) training algorithm which is simple and relatively powerful. The ML training is an efficient algorithm to individually estimate the model parameters of each class under the assumption that all class models in a classifier are statistically independent. That assumption, however, is not valid in many real situations, which degrades the performance of the classifier. In this paper, we propose a minimum-error-rate training algorithm based on the MAP (Maximum a Posteriori) approach. The algorithm regards the normalized outputs of the classifier as estimates of the a posteriori probability, and tries to maximize those estimates. According to Bayes decision theory, the proposed algorithm satisfies the condition of minimum-error-rate classificatin. We apply this algorithm to NPM (Neural Prediction Model) for speech recognition, and derive new disrminative training algorithms. Experimental results on ten Korean digits recognition have shown the reduction of 37.5% of the number of recognition errors.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2001.05a
• /
• pp.105-108
• /
• 2001

In this paper, the learning algorithm called advanced expanded and truncate algorithm(AETLA) is proposed to training multilayer binary neural network to approximate binary to binary mapping. AETLA used merit of ETL and MTGA learning algorithm. We proposed to new learning algorithm to decrease number of hidden layer. Therefore, learning speed of the proposed AETLA learning algorithm is much faster than other learning algorithm.

• Journal of Information Processing Systems
• /
• v.14 no.5
• /
• pp.1075-1086
• /
• 2018

In this paper a cross-validation algorithm for training probabilistic neural networks (PNNs) is presented in order to be applied to automatic face identification. Actually, standard PNNs perform pretty well for small and medium sized databases but they suffer from serious problems when it comes to using them with large databases like those encountered in biometrics applications. To address this issue, we proposed in this work a new training algorithm for PNNs to reduce the hidden layer's size and avoid over-fitting at the same time. The proposed training algorithm generates networks with a smaller hidden layer which contains only representative examples in the training data set. Moreover, adding new classes or samples after training does not require retraining, which is one of the main characteristics of this solution. Results presented in this work show a great improvement both in the processing speed and generalization of the proposed classifier. This improvement is mainly caused by reducing significantly the size of the hidden layer.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.11
• /
• pp.158-165
• /
• 1999

It has been established that a crack has an important effect on the dynamic behavior of a structure. This effect depends mainly on the location and depth of the crack. To identify the location and depth of a crack in a structure, a method is presented in this paper which uses hybrid neuro-genetic technique. Feed-forward multilayer neural networks trained by back-propagation are used to learn the input)the location and dept of a crack)-output(the structural eigenfrequencies) relation of the structural system. With this neural network and genetic algorithm, it is possible to formulate the inverse problem. Neural network training algorithm is the back propagation algorithm with the momentum method to attain stable convergence in the training process and with the adaptive learning rate method to speed up convergence. Finally, genetic algorithm is used to fine the minimum square error.

• Journal of Information Processing Systems
• /
• v.14 no.6
• /
• pp.1464-1479
• /
• 2018

In face recognition, sometimes the number of available training samples for single category is insufficient. Therefore, the performances of models trained by convolutional neural network are not ideal. The small sample face recognition algorithm based on novel Siamese network is proposed in this paper, which doesn't need rich samples for training. The algorithm designs and realizes a new Siamese network model, SiameseFacel, which uses pairs of face images as inputs and maps them to target space so that the $L_2$ norm distance in target space can represent the semantic distance in input space. The mapping is represented by the neural network in supervised learning. Moreover, a more lightweight Siamese network model, SiameseFace2, is designed to reduce the network parameters without losing accuracy. We also present a new method to generate training data and expand the number of training samples for single category in AR and labeled faces in the wild (LFW) datasets, which improves the recognition accuracy of the models. Four loss functions are adopted to carry out experiments on AR and LFW datasets. The results show that the contrastive loss function combined with new Siamese network model in this paper can effectively improve the accuracy of face recognition.

• Proceedings of the KIPE Conference
• /
• 1998.10a
• /
• pp.293-297
• /
• 1998

This paper presents a newly developed speed sensorless drive using RLS based on Neural Network Training Algorithm. The proposed algorithm has just the time-varying learning rate, while the wellknown back-propagation algorithm based on gradient descent has a constant learning rate. The number of iterations required by the new algorithm to converge is less than that of the back-propagation algorithm. The theoretical analysis and experimental results to verify the effectiveness of the proposed control strategy are described.