• Journal of Digital Contents Society
• /
• v.18 no.4
• /
• pp.739-745
• /
• 2017

This paper presents and discusses an implementation of the GPU shifted sorting method to find approximate k nearest neighbors which executes within "warp", the minimum execution unit in GPU parallel architecture. Also, this paper presents the comparison results with other two common nearest neighbor searching methods, GPU-based kd-tree and ANN (Approximate Nearest Neighbor) library. The proposed implementation focuses on the cases when k is small, i.e. 2, 4, 8, and 16, which are handled efficiently within warp to consider it is very common for applications to handle small k's. Also, this paper discusses optimization ways to implementation by improving memory management in a loop for the CUB open library and adopting CUDA commands which are supported by GPU hardware. The proposed implementation shows more than 16-fold speed-up against GPU-based other methods in the tests, implying that the improvement would become higher for more larger input data.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.43 no.6
• /
• pp.883-896
• /
• 2023

As human activities in Korea are spread throughout the mountains, forest fires often affect residential areas, infrastructure, and other facilities. Hence, it is necessary to detect fire-damaged areas quickly to enable support and recovery. Remote sensing is the most efficient tool for this purpose. Fire damage detection experiments were conducted on the east coast of Korea. Because this area comprises a mixture of forest and artificial land cover, data with low resolution are not suitable. We used Sentinel-2 multispectral instrument (MSI) data, which provide adequate temporal and spatial resolution, and the k-nearest neighbor (kNN) algorithm in this study. Six bands of Sentinel-2 MSI and two indices of normalized difference vegetation index (NDVI) and normalized burn ratio (NBR) were used as features for kNN classification. The kNN classifier was trained using 2,000 randomly selected samples in the fire-damaged and undamaged areas. Outliers were removed and a forest type map was used to improve classification performance. Numerous experiments for various neighbors for kNN and feature combinations have been conducted using bi-temporal and uni-temporal approaches. The bi-temporal classification performed better than the uni-temporal classification. However, the uni-temporal classification was able to detect severely damaged areas.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2005.05a
• /
• pp.423-426
• /
• 2005

기억기반학습의 일종인 최근접 이웃(k nearest neighbor) 알고리즘은 과거의 데이터들 중에서 새로운 개체와 유사한 데이터들을 이용해서 새로운 개체의 목적 값을 예측하는 것이다. 이 경우 속성의 가중치를 계산하는 방식은 kNN의 성능을 결정하는 중요한 요소가 된다. 본 논문에서는 기존의 다른 이론들과 달리 정보이론에서 사용되는 엔트로피 개념을 이용해서 속성의 가중치를 이론적이고, 효과적으로 계산하는 새로운 방법을 제시하고자한다. 제안된 방법은 각 속성이 목적속성에 제공하는 정보의 양에 따라 가중치를 자동으로 계산하여 kNN의 성능을 향상시킨다. 마지막으로 이러한 방식의 성능을 다수의 실험을 통해 비교하였다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2008.06c
• /
• pp.376-380
• /
• 2008

사회의 복잡화와 인터넷의 성장으로 폭발적으로 늘어나고 있는 정보들을 사용자가 모두 검토한 후 여과하기는 어려운 일이다. 이러한 문제를 보완하기 위해서 자동화된 정보 여과 기술이 사용되는데, k-최근접 이웃(k-nearest neighbor) 알고리즘은 그 구현이 간단하며 비교적 정확하여 가장 널리 쓰이고 있는 알고리즘 중 하나이다. k 개의 최근접 이웃들로부터 평가값을 계산하는 데 흔히 쓰이는 방법은 상관계수를 이용한 가중치에 기반하는 것이다. 본 논문에서는 이를 보완하여 대규모 데이터에 대해서도 속도는 크게 저하되지 않으며 정확도는 대폭 향상시킬 수 있는 방법을 적용하였다. 또한, 최근접 이웃을 구하는 거리함수로 다양한 방법을 시도하였다. 영화추천을 위한 실제 데이터에 대한 실험 결과, 속도의 저하는 미미하였으나 정확도에 있어서는 크게 향상된 결과를 가져올 수 있었다.

• Proceedings of the KIPE Conference
• /
• 2005.07a
• /
• pp.624-627
• /
• 2005

This paper presents a new data classification method using the Robbins Monro stochastic approximation algorithm k-nearest neighbor and distribution analysis. To cluster the data set, we decide the centroid of the test data set using k-nearest neighbor algorithm and the local area of data set. To decide each class of the data, the Robbins Monro stochastic approximation algorithm is applied to the decided local area of the data set. To evaluate the performance, the proposed classification method is compared to the conventional fuzzy c-mean method and k-nn algorithm. The simulation results show that the proposed method is more accurate than fuzzy c-mean method, k-nn algorithm and discriminant analysis algorithm.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.10a
• /
• pp.379-383
• /
• 2005

This paper proposes a new steel strip surface inspection system. The system acquires bright and dark field images of defects by using a stroboscopic IR LED light and area camera system and the defect images are preprocessed and segmented in real time for feature extraction. 4113 defect samples of cold roll steel strips are used to develop KNN (k-Nearest Neighbor) classifier which classifies the defects into 8 different types. The developed KNN classifier demonstrates about 85% classifying performance which is considered very plausible result.

• ETRI Journal
• /
• v.40 no.4
• /
• pp.531-536
• /
• 2018

The weighted K-nearest neighbor (WKNN) algorithm is used to reduce positioning accuracy, as it uses a fixed number of neighbors to estimate the position. In this paper, we propose a dynamic threshold location algorithm (DH-KNN) to improve positioning accuracy. The proposed algorithm is designed based on a dynamic threshold to determine the number of neighbors and filter out singular reference points (RPs). We compare its performance with the WKNN and Enhanced K-Nearest Neighbor (EKNN) algorithms in test spaces of networks with dimensions of $20m{\times}20m$, $30m{\times}30m$, $40m{\times}40m$ and $50m{\times}50m$. Simulation results show that the maximum position accuracy of DH-KNN improves by 31.1%, and its maximum position error decreases by 23.5%. The results demonstrate that our proposed method achieves better performance than other well-known algorithms.

• International Journal of Contents
• /
• v.1 no.1
• /
• pp.50-58
• /
• 2005

Generally, multidimensional data such as image and spatial data require large amount of storage space. There is a limit to store and manage those large amounts of data in single workstation. If we manage the data on parallel computing environment which is being actively researched these days, we can get highly improved performance. In this paper, we propose a parallel multidimensional index structure that exploits the parallelism of the parallel computing environment. The proposed index structure is nP(processor)-nxmD(disk) architecture which is the hybrid type of nP-nD and 1P-nD. Its node structure in-creases fan-out and reduces the height of an index. Also, a range search algorithm that maximizes I/O parallelism is devised, and it is applied to k-nearest neighbor queries. Through various experiments, it is shown that the proposed method outperforms other parallel index structures.

• Proceedings of the IEEK Conference
• /
• 2008.06a
• /
• pp.833-834
• /
• 2008

In this paper, a pattern classification method using closest decision method based on the mean of norm in the closet prototype from an input pattern and its k nearest neighbor prototypes is presented to do accurate classification in arbitrary distributed patterns when the number of patterns is very low. Also this method can be used to classify input pattern precisely when the number patterns is very low because this method considers the weight by the difference of variance in prototypes around the discrimination boundary.

• Journal of the Korea Society of Computer and Information
• /
• v.10 no.2 s.34
• /
• pp.215-221
• /
• 2005

TRecently, there has been enormous growth in the amount of commercial and scientific data, such as protein sequences, retail transactions, and web-logs. Such datasets consist of sequence data that have an inherent sequential nature. In this Paper, we study how to classify these sequence datasets. There are several kinds techniques for data classification such as decision tree induction, Bayesian classification and K-NN etc. In our approach, we use a K-NN algorithm for classifying sequences. In addition, we propose a new similarity measure to compute the similarity between two sequences and an efficient method for measuring similarity.