• Journal of Korea Game Society
• /
• v.20 no.5
• /
• pp.53-64
• /
• 2020

Supporting NNS method in KD-Tree algorithm is essential in multidimensional data applications. In this paper, we propose HD-Tree, a high-performance Lock-Free KD-Tree that supports NNS in situations where reads and writes occurs concurrently. HD-Tree reduced the number of synchronization nodes used in NNS and requires less atomic operations during Lock-Free method execution. Comparing with existing algorithms, in a multi-core system with 8 core 16 thread, HD-Tree's performance has improved up to 95% on NNS and 15% on modifying in oversubscription situation.

• The Journal of the Korea Contents Association
• /
• v.14 no.10
• /
• pp.454-462
• /
• 2014

With the development of location aware technologies and mobile devices, location-based services have been studied. To provide location-based services, many researchers proposed methods for processing various query types with Mapreduce(MR). One of the proposed methods, is a Reverse k-nearest neighbor(RkNN) query processing method with MR. However, the existing methods spend too much cost to process the continuous RkNN query. In this paper, we propose an efficient continuous RkNN query processing method with MR to resolve the problems of the existing methods. The proposed method uses the 60-degree-pruning method. The proposed method does not need to reprocess the query for continuous query processing because the proposed method draws and monitors the monitoring area including the candidate objects of a RkNN query. In order to show the superiority of the proposed method, we compare it with the query processing performance of the existing method.

• Journal of Korean Society for Geospatial Information Science
• /
• v.18 no.4
• /
• pp.119-129
• /
• 2010

The joint use of remotely sensed data and field measurements has been widely used to estimate aboveground carbon stock in many countries. Recently, Korea Forest Research Institute has developed new carbon emission factors for kind of tree, thus more accurate estimate is possible. In this study, the aboveground carbon stock of Danyang area in South Korea was estimated using k-Nearest Neighbor(kNN) algorithm with the 5th National Forest Inventory(NFI) data. Considering the spectral response of forested area under the climate condition in Korea peninsular which has 4 distinct seasons, Landsat TM seasonal satellite images were collected. As a result, the estimated total carbon stock of Danyang area was ranged from 3542768.49tonC to 3329037.51tonC but seasonal trends were not found.

• Journal of Korea Spatial Information System Society
• /
• v.11 no.3
• /
• pp.9-18
• /
• 2009

Recently, due to the advanced technologies of mobile devices and wireless communication, there are many studies on telematics and LBS(location-based service) applications. because moving objects usually move on spatial networks, their locations are updated frequently, leading to the degradation of retrieval performance. To manage the frequent updates of moving objects' locations in an efficient way, a new distributed grid scheme, called DS-GRID (distributed S-GRID), and k-NN(k-nearest neighbor) query processing algorithm was proposed[1]. However, the result of k-NN query processing technique may be invalidated as the location of query and moving objects are changed. Therefore, it is necessary to study on continuous k-NN query processing algorithm. In this paper, we propose both MCE-CKNN and MBP(Monitoring in Border Point)-CKNN algorithmss are S-GRID. The MCE-CKNN algorithm splits a query route into sub-routes based on cell and seproves retrieval performance by processing query in parallel way by. In addition, the MBP-CKNN algorithm stores POIs from the border points of each grid cells and seproves retrieval performance by decreasing the number of accesses to the adjacent cells. Finally, it is shown from the performance analysis that our CKNN algorithms achieves 15-53% better retrieval performance than the Kolahdouzan's algorithm.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10b
• /
• pp.145-147
• /
• 2001

문서 자동 분류에는 통계적인 기법과 machine learning 기법의 맡은 알고리즘들이 이용되고 있다. 통계적인 기법 알고리즘을 이용한 문서 분류는 높은 성능을 보이지만 분류할 카테고리가 둘 이상인 경우가 빈번할 경우에는 정확률이 급격히 저하되는 단점이 있다. 본 논문에서는 K-NN알고리즘을 이용하여 일차적인 문서 분류를 수행한 후 특정 카테고리로 분류하기에 애매모호한 경우가 생길 경우 시소러스의 일반화 관계와 연관화 관계를 이용하여 모호성을 줄임으로써 문서 자동 분류의 성능을 높이기 위한 새 기법을 제안한다.

• Journal of Korea Spatial Information System Society
• /
• v.10 no.4
• /
• pp.11-19
• /
• 2008

Recently, advances in mobile devices and wireless communication technologies require research on various location-based services. As a result, many studies on processing k-nearest neighbor query, which is most im portant one in location-based services, have been done. Most of existing studies use pre-computation technique to improve retrieval performance by computing network distance between POIs and nodes beforehand in spatial networks. However, they have a drawback that they can not deal with effectively the update of POIs to be searched. In this paper, we propose a distributed grid scheme using S-GRID to overcome the disadvantage of the existing work as well as to manage the location information of a large number of moving objects in efficient way. In addition, we describe a k-nearest neighbor(k-NN) query processing algorithm for the proposed distributed grid scheme. Finally, we show the efficiency of our distributed grid scheme by making a performance comparison between the k-NN query processing algorithm of our scheme and that of S-GRID.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.4
• /
• pp.273-280
• /
• 2013

This paper presents novel OCS-LBP (Oriented Center Symmetric Local Binary Patterns) based on orientation of pixel gradient and image retrieval system based on BoF (Bag-of-Feature) and random forest classifier. Feature vectors extracted from training data are clustered into code book and each feature is transformed new BoF feature using code book. BoF features are applied to random forest for training and random forest having N classes is constructed by combining several decision trees. For testing, the same OCS-LBP feature is extracted from a query image and BoF is applied to trained random forest classifier. In contrast to conventional retrieval system, query image selects similar K-nearest neighbor (K-NN) classes after random forest is performed. Then, Top K similar images are retrieved from database images that are only labeled K-NN classes. Compared with other retrieval algorithms, the proposed method shows both fast processing time and improved retrieval performance.

• Journal of Korea Multimedia Society
• /
• v.14 no.6
• /
• pp.711-718
• /
• 2011

The k-Nearest Neighbor classifier that does not require a training phase is appropriate for a variable number of classes problem like face recognition, Recently distance metric learning methods that is trained with a given data set have reported the significant improvement of the kNN classifier. However, the performance of a distance metric learning method is variable for each application, In this paper, we focus on the face recognition and compare the performance of the state-of-the-art distance metric learning methods, Our experimental results on the public face databases demonstrate that the Mahalanobis distance metric based on PCA is still competitive with respect to both performance and time complexity in face recognition.

• Journal of Korean Society for Quality Management
• /
• v.47 no.4
• /
• pp.687-700
• /
• 2019

Purpose: The purpose of this study is to investigate the interaction effects between price determinants of artworks. We expand the methodology in art market by applying machine learning techniques to estimate the price of artworks and compare linear regression and machine learning in terms of prediction accuracy. Methods: Moderated regression analysis was performed to verify the interaction effects of artistic characteristics on price. The moderating effects were studied by confirming the significance level of the interaction terms of the derived regression equation. In order to derive price estimation model, we use multiple linear regression analysis, which is a parametric statistical technique, and k-nearest neighbor (kNN) regression, which is a nonparametric statistical technique in machine learning methods. Results: Mostly, the influences of the price determinants of art are different according to the auction types and the artist 's reputation. However, the auction type did not control the influence of the genre of the work on the price. As a result of the analysis, the kNN regression was superior to the linear regression analysis based on the prediction accuracy. Conclusion: It provides a theoretical basis for the complexity that exists between pricing determinant factors of artworks. In addition, the nonparametric models and machine learning techniques as well as existing parameter models are implemented to estimate the artworks' price.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.2
• /
• pp.742-762
• /
• 2015

In this study, we propose a distance metric learning approach called discriminant metric learning (DML) for face verification, which addresses a binary-class problem for classifying whether or not two input images are of the same subject. The critical issue for solving this problem is determining the method to be used for measuring the distance between two images. Among various methods, the large margin nearest neighbor (LMNN) method is a state-of-the-art algorithm. However, to compensate the LMNN's entangled data distribution due to high levels of appearance variations in unconstrained environments, DML's goal is to penalize violations of the negative pair distance relationship, i.e., the images with different labels, while being integrated with LMNN to model the distance relation between positive pairs, i.e., the images with the same label. The likelihoods of the input images, estimated using DML and LMNN metrics, are then weighted and combined for further analysis. Additionally, rather than using the k-nearest neighbor (k-NN) classification mechanism, we propose a verification mechanism that measures the correlation of the class label distribution of neighbors to reduce the false negative rate of positive pairs. From the experimental results, we see that DML can modify the relation of negative pairs in the original LMNN space and compensate for LMNN's performance on faces with large variances, such as pose and expression.