• ETRI Journal
• /
• v.22 no.3
• /
• pp.10-18
• /
• 2000

Although an n-th order cross-entropy (nCE) error function resolves the incorrect saturation problem of conventional error backpropagation (EBP) algorithm, performance of multilayer perceptrons (MLPs) trained using the nCE function depends heavily on the order of nCE. In this paper, we propose an adaptive learning rate to markedly reduce the sensitivity of MLP performance to the order of nCE. Additionally, we propose to limit error signal values at out-put nodes for stable learning with the adaptive learning rate. Through simulations of handwritten digit recognition and isolated-word recognition tasks, it was verified that the proposed method successfully reduced the performance dependency of MLPs on the nCE order while maintaining advantages of the nCE function.

• Journal of information and communication convergence engineering
• /
• v.2 no.1
• /
• pp.36-39
• /
• 2004

In this paper, a method of improving the learning speed and convergence rate is proposed to exploit the advantages of artificial neural networks and neuro-fuzzy systems. This method is applied to the XOR problem, n bit parity problem, which is used as the benchmark in the field of pattern recognition. The method is also applied to the recognition of digital image for practical image application. As a result of experiment, it does not always guarantee convergence. However, the network showed considerable improvement in learning time and has a high convergence rate. The proposed network can be extended to any number of layers. When we consider only the case of the single layer, the networks had the capability of high speed during the learning process and rapid processing on huge images.

• Proceedings of the KIEE Conference
• /
• 1997.07b
• /
• pp.386-389
• /
• 1997

Error-backpropagation has been used in the bulk of Practical applications for neural networks. While an emulator, a multilayered neural network, learns to identify the system's dynamic characteristics. There is, however, no concrete theoretical results about the structure of a plant and the structure of a multilayered neural network and the learning rate. The paper investigates the relation between structure of a plant and a multilayered network and learning rate. Simulation study shows that the plant signal with a short period and a fast sam time is preferable for learning of the network emulator.

• Journal of the Semiconductor & Display Technology
• /
• v.22 no.2
• /
• pp.40-44
• /
• 2023

The defect detection based on deep learning requires minimal loss and high accuracy to pinpoint product defects. In this paper, we confirm the loss rate of deep learning training based on disc-shaped artificial scaffold images. It is intended to compare the performance of Cross-Entropy functions used in object detection algorithms. The model was constructed using normal, defective artificial scaffold images and category cross entropy and sparse category cross entropy. The data was repeatedly learned five times using each loss function. The average loss rate, average accuracy, final loss rate, and final accuracy according to the loss function were confirmed.

• Journal of information and communication convergence engineering
• /
• v.9 no.6
• /
• pp.637-640
• /
• 2011

The hybrid neural networks have characteristics such as fast learning times, generality, and simplicity, and are mainly used to classify learning data and to model non-linear systems. The middle layer of a hybrid neural network clusters the learning vectors by grouping homogenous vectors in the same cluster. In the clustering procedure, the homogeneity between learning vectors is represented as the distance between the vectors. Therefore, if the distances between a learning vector and all vectors in a cluster are smaller than a given constant radius, the learning vector is added to the cluster. However, the usage of a constant radius in clustering is the primary source of errors and therefore decreases the recognition success rate. To improve the recognition success rate, we proposed the enhanced hybrid network that organizes the middle layer effectively by using the enhanced ART1 network adjusting the vigilance parameter dynamically according to the similarity between patterns. The results of experiments on a large number of calling card images showed that the proposed algorithm greatly improves the character extraction and recognition compared with conventional recognition algorithms.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.1
• /
• pp.451-465
• /
• 2017

This paper presents a novel discriminative visual tracking algorithm with an adaptive incremental extreme learning machine. The parameters for an adaptive incremental extreme learning machine are initialized at the first frame with a target that is manually assigned. At each frame, the training samples are collected and random Haar-like features are extracted. The proposed tracker updates the overall output weights for each frame, and the updated tracker is used to estimate the new location of the target in the next frame. The adaptive learning rate for the update of the overall output weights is estimated by using the confidence of the predicted target location at the current frame. Our experimental results indicate that the proposed tracker can manage various difficulties and can achieve better performance than other state-of-the-art trackers.

• Journal of IKEEE
• /
• v.23 no.1
• /
• pp.112-119
• /
• 2019

In this paper, we propose a deep learning structure suitable for embedded system. The flame detection process of the proposed deep learning structure consists of four steps : flame area detection using flame color model, flame image classification using deep learning structure for flame color specialization, $N{\times}N$ cell separation in detected flame area, flame image classification using deep learning structure for flame shape specialization. First, only the color of the flame is extracted from the input image and then labeled to detect the flame area. Second, area of flame detected is the input of a deep learning structure specialized in flame color and is classified as flame image only if the probability of flame class at the output is greater than 75%. Third, divide the detected flame region of the images classified as flame images less than 75% in the preceding section into $N{\times}N$ units. Fourthly, small cells divided into $N{\times}N$ units are inserted into the input of a deep learning structure specialized to the shape of the flame and each cell is judged to be flame proof and classified as flame images if more than 50% of cells are classified as flame images. To verify the effectiveness of the proposed deep learning structure, we experimented with a flame database of ImageNet. Experimental results show that the proposed deep learning structure has an average resource occupancy rate of 29.86% and an 8 second fast flame detection time. The flame detection rate averaged 0.95% lower compared to the existing deep learning structure, but this was the result of light construction of the deep learning structure for application to embedded systems. Therefore, the deep learning structure for flame detection proposed in this paper has been proved suitable for the application of embedded system.

• 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.

• Educational Technology International
• /
• v.6 no.1
• /
• pp.41-63
• /
• 2005

E-Learning is widely introduced with cyber universities in Korea from 2001 whencyber universities were first authorized by the Ministry of Education and Human Resource Development. E-learning amplified by cyber university gave a big impact in the campus based university which became the cause for the educational paradigm shift. The changes of status of cyber university shows important trend in college education which was analyzed by enrollment rate, types of cyber university, demography, and study areas. The enrollment rate of cyber universities is ever since 2001 and variety of study areas gives popularity to students. The demography of students is as expected older than traditional students. Female students at the cyber university outnumbered that at campus based university in Korea. For analyzing the trend of e-learning in Korea, there were studies twice in 2001 May-June from 213 faculty members and staff, 630 students and in 2004 May-June with 401 students. Most of e-learning students tent to spend less time yet, students feel more burden with e-learning. Professors tend to load more materials for the e-learning in 2001but in 2004 study, the difference no longer exists. Professors and students feel the academic achievement through e-learning is not as good as from the traditional classes. Difficulties for e-learning in 2001 were the lack of administrative information but in 2004, boring contents and lack of instructional strategies for e-learning. Technical problems still do exist but less serious. Suggestions for e-learning are blended learning, online students prefer video streaming with their own lecturer, new definition of instructor is needed, professional development for content development and online instruction is needed, success story of online learning should be encouraged, guidance for online students needed. The cyber university experiencegave a positive impact on the traditional universities such as rethinking the roles of universities, the quality control of classes, professional development, student oriented educational service of e-learning pedagogy.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.9
• /
• pp.1682-1688
• /
• 2008

The Counterpropagation algorithm(CP) is a combination of Kohonen competition network as a hidden layer and the outstar structure of Grossberg as an output layer. CP has been used in many real applications for pattern matching, classification, data compression and statistical analysis since its learning speed is faster than other network models. However, due to the Kohonen layer's winner-takes-all strategy, it often causes instable learning and/or incorrect pattern classification when patterns are relatively diverse. Also, it is often criticized by the sensitivity of performance on the learning rate. In this paper, we propose an enhanced CP that has multiple Kohonen layers and dynamic controlling facility of learning rate using the frequency of winner neurons and the difference between input vector and the representative of winner neurons for stable learning and momentum learning for controlling weights of output links. A real world application experiment - pattern recognition from passport information - is designed for the performance evaluation of this enhanced CP and it shows that our proposed algorithm improves the conventional CP in learning and recognition performance.