• International Journal of Computer Science & Network Security
• /
• v.23 no.6
• /
• pp.99-106
• /
• 2023

Now a days, large volumes of data is accumulating in every field due to increase in capacity of storage devices. These large volumes of data can be applied with data mining for finding useful patterns which can be used for business growth, improving services, improving health conditions etc. Data from different sources can be combined before applying data mining. The data thus gathered can be misused for identity theft, fake credit/debit card transactions, etc. To overcome this, data mining techniques which provide privacy are required. There are several privacy preserving data mining techniques available in literature like randomization, perturbation, anonymization etc. This paper proposes an Enhanced Hybrid Privacy Preserving Data Mining(EHPPDM) technique. The proposed technique provides more privacy of data than existing techniques while providing better classification accuracy. The experimental results show that classification accuracies have increased using EHPPDM technique.

• Journal of Information Processing Systems
• /
• v.20 no.4
• /
• pp.501-513
• /
• 2024

In recent years, graph neural networks (GNNs) have been extensively used to analyze graph data across various domains because of their powerful capabilities in learning complex graph-structured data. However, recent research has focused on improving the performance of a single GNN with only two or three layers. This is because stacking layers deeply causes the over-smoothing problem of GNNs, which degrades the performance of GNNs significantly. On the other hand, ensemble methods combine individual weak models to obtain better generalization performance. Among them, gradient boosting is a powerful supervised learning algorithm that adds new weak models in the direction of reducing the errors of the previously created weak models. After repeating this process, gradient boosting combines the weak models to produce a strong model with better performance. Until now, most studies on GNNs have focused on improving the performance of a single GNN. In contrast, improving the performance of GNNs using multiple GNNs has not been studied much yet. In this paper, we propose gradient boosted graph neural networks (GBGNN) that combine multiple shallow GNNs with gradient boosting. We use shallow GNNs as weak models and create new weak models using the proposed gradient boosting-based loss function. Our empirical evaluations on three real-world datasets demonstrate that GBGNN performs much better than a single GNN. Specifically, in our experiments using graph convolutional network (GCN) and graph attention network (GAT) as weak models on the Cora dataset, GBGNN achieves performance improvements of 12.3%p and 6.1%p in node classification accuracy compared to a single GCN and a single GAT, respectively.

• Journal of the Korean Society of Industry Convergence
• /
• v.17 no.1
• /
• pp.27-34
• /
• 2014

This paper proposes a new approach to the designed of visual feedback control system based on visual servoing method. The main focus of this paper is presented how it is effective to use many features for improving the accuracy of the visual feedback control of industrial articulated robot for assembling and inspection of parts. Some rank conditions, which relate the image Jacobian, and the control performance are derived. It is also proven that the accuracy is improved by increasing the number of features. The effectiveness of redundant features is verified by the real time experiments on a SCARA type robot(FARA) made in samsung electronics company.

• Journal of the Korean Society of Combustion
• /
• v.18 no.1
• /
• pp.1-6
• /
• 2013

Authors' previous works on thermoacoustic(TA) model development showed good results in predicting combustion instability characteristics in a gas turbine combustor. However, they also suggested there were some limitations in growth rate estimation, which might be related with over-simplification of flame structure. As a first trial for improving the model accuracy, the current paper introduces the modified TA model considering the actual flame location in the combustor. The combustor is divided into the unburned and the burned area before and after the flame location, and then acoustic equations are re-organized. The modified TA model results show a better accuracy in predicting the growth rate of instabilities comparing with the previous results. However, obtained results still overestimate the conditions where the combustor goes unstable. Further researches considering heat release distribution through flames are required.

• ETRI Journal
• /
• v.37 no.2
• /
• pp.251-261
• /
• 2015

We present a novel method for real-time automatic license plate detection in high-resolution videos. Although there have been extensive studies of license plate detection since the 1970s, the suggested approaches resulting from such studies have difficulties in processing high-resolution imagery in real-time. Herein, we propose a novel cascade structure, the fastest classifier available, by rejecting false positives most efficiently. Furthermore, we train the classifier using the core patterns of various types of license plates, improving both the computation load and the accuracy of license plate detection. To show its superiority, our approach is compared with other state-of-the-art approaches. In addition, we collected 20,000 images including license plates from real traffic scenes for comprehensive experiments. The results show that our proposed approach significantly reduces the computational load in comparison to the other state-of-the-art approaches, with comparable performance accuracy.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2000.04b
• /
• pp.159-167
• /
• 2000

In optimization problems, computationally intensive or expensive simulations hinder the use of standard optimization techniques because the computational expense is too heavy to implement them at each iteration of the optimization algorithm. Therefore, those expensive simulations are often replaced with approximation models which can be evaluated nearly free. However, because of the limited accuracy of the approximation models, it is practically impossible to find an exact optimal point of the original problem. Significant efforts have been made to overcome this problem. The approximation models are sequentially updated during the iterative optimization process such that interesting design points are included. The interesting points have a strong influence on making the approximation model capture an overall trend of the original function or improving the accuracy of the approximation in the vicinity of a minimizer. They are successively determined at each iteration by utilizing the predictive ability of the approximation model. This paper will focuses on those approaches and introduces various approximation methods.

• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.977-985
• /
• 2020

Aiming at the problem of complex feature extraction and low accuracy in human action recognition, this paper proposed a network structure combining batch normalization algorithm with GoogLeNet network model. Applying Batch Normalization idea in the field of image classification to action recognition field, it improved the algorithm by normalizing the network input training sample by mini-batch. For convolutional network, RGB image was the spatial input, and stacked optical flows was the temporal input. Then, it fused the spatio-temporal networks to get the final action recognition result. It trained and evaluated the architecture on the standard video actions benchmarks of UCF101 and HMDB51, which achieved the accuracy of 93.42% and 67.82%. The results show that the improved convolutional neural network has a significant improvement in improving the recognition rate and has obvious advantages in action recognition.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.773-776
• /
• 2005

In this paper, For improving the positioning accuracy of hydrostatic table, relationship between temperature of atmosphere and thermal characteristics of hydrostatic table is analyzed, and influence of thermal characteristics on positioning accuracy is also analyzed experimentally. From the experimental results, it is confirmed that positioning error and repeatability is $0.21{\mu}m\;and\;0.18{\mu}m$ when the laser scale which has $0.01{\mu}m$ of resolution is used as feed-back unit. and also confirmed that thermal deformation of scale and supporter, which occurs by the temperature variation of atmosphere, works as limit of repeatability in long time operation.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2002.05a
• /
• pp.37-41
• /
• 2002

Thermal characteristics of hydrostatic guideway is largely depended on the temperature of supplied oil. For improving the positioning accuracy of hydrostatic guideway, relationship between setting temperature of oil cooler and thermal characteristics is analyzed, and influence of thermal characteristics on positioning accuracy is also analyzed experimental1y in this paper. Laser scale which has 0.01 $\mu\textrm{m}$ of resolution is used as feed-back unit. From the experimental results, it is confirmed that positioning error and repeatability is minimize upto 0.21 $\mu\textrm{m}$ and 0.18 $\mu\textrm{m}$ when the temperature of supplied oil is setting equal to temperature of atmosphere, and also confirmed that thermal deformation, which occurs by the temperature deviation between table and rail or scale supporter, works as limit of repeatability in long time operation.

• Proceedings of the Korea Institutes of Information Security and Cryptology Conference
• /
• 2006.06a
• /
• pp.133-137
• /
• 2006

In iris recognition, iris segmentation plays a very important role because its accuracy affects directly to the performance of the whole system. This paper proposes a new approach for segmenting iris that is fast, accurate and especially robust to occlusion and illumination. In this method, a circular filter is used for detecting the center of the inner circle. Then, a technique to linearize the limbus is applied and the limbus is detected using a linear filter. Experimental results show that the proposed method has promising performance for improving the iris recognition accuracy.