• Proceedings of the Korea Multimedia Society Conference
• /
• 2001.06a
• /
• pp.116-119
• /
• 2001

본 논문에서는 주요성분분석 기법을 도입하여 회귀분석을 위한 신경망의 성능 개선방안을 제안하였다. 이는 대용량의 입력 데이터를 통계적으로 독립인 특징들의 집합으로 변환시키는 주요성분분석 기법의 속성을 살려 학습데이터의 타원을 감소시킴으로서 고차원의 학습데이터에 따른 신경망의 학습성능 의존성을 줄이기 위함이다. 제안된 기법의 신경망을 10개의 독립변수 패턴을 가진 자동차 연비문제에 적용하여 시뮬레이션한 결과, 기존의 학습데이터를 그대로 이용하는 신경 망보다 우수한 학습성능과 회귀성능이 있음을 확인할 수 있었다.

• Journal of KIISE:Databases
• /
• v.37 no.2
• /
• pp.71-81
• /
• 2010

The problem of finding clusters is widely used in numerous applications, such as pattern recognition, image analysis, market analysis. The important factors that decide cluster quality are the similarity measure and the number of attributes. Similarity measures should be defined with respect to the data types. Existing similarity measures are well applicable to numerical attribute values. However, those measures do not work well when the data is described by categorical attributes, that is, when no inherent similarity measure between values. In high dimensional spaces, conventional clustering algorithms tend to break down because of sparsity of data points. To overcome this difficulty, a subspace clustering approach has been proposed. It is based on the observation that different clusters may exist in different subspaces. In this paper, we propose a new similarity measure for clustering of high dimensional categorical data. The measure is defined based on the fact that a good clustering is one where each cluster should have certain information that can distinguish it with other clusters. We also try to capture on the attribute dependencies. This study is meaningful because there has been no method to use both of them. Experimental results on real datasets show clusters obtained by our proposed similarity measure are good enough with respect to clustering accuracy.

• The KIPS Transactions:PartD
• /
• v.8D no.2
• /
• pp.145-153
• /
• 2001

The Pyramid-Technique is based on mapping n-dimensional space data into one-dimensional data and expressing it as B-tree ; and by solving the problem of search time complexity the pyramid technique also prevents the effect \"phenomenon of dimensional curse\" which is caused by treatment of hypercube range query in n-dimensional data space. The Spherical Pyramid-Technique applies the pyramid method’s space division strategy, uses spherical range query and improves the search performance to make it suitable for similarity search. However, depending on the size of data and change in dimensions, the two above technique demonstrate significantly inferior search performance for data sizes greater than one million and dimensions greater than sixteen. In this paper, we propose a new index-structured PdR-Tree to improve the search performance for high dimensional data such as multimedia data. Test results using simulation data as well as real data demonstrate that PdR-Tree surpasses both the Pyramid-Technique and Spherical Pyramid-Technique in terms of search performance.

• The KIPS Transactions:PartD
• /
• v.11D no.2
• /
• pp.269-280
• /
• 2004

The Pyramid-Technique is based on mapping n-dimensional space data into one-dimensional data and expresses it as a B+-tree. By solving the problem of search time complexity the pyramid technique also prevents the effect of "phenomenon of dimensional curse" which is caused by treatment of hypercube range query in n-dimensional data space. The SPY-TEC applies the space division strategy in pyramid method and uses spherical range query suitable for similarity search so that Improves the search performance. However, nearest neighbor query is more efficient than range query because it is difficult to specify range in similarity search. Previously proposed index methods perform well only in the specific distribution of data. In this paper, we propose an efficient searching technique for nearest neighbor object using PdR-Tree suggested to improve the search performance for high dimensional data such as multimedia data. Test results, which uses simulation data with various distribution as well as real data, demonstrate that PdR-Tree surpasses both the Pyramid-Technique and SPY-TEC in views of search performance.rformance.

• The Korean Journal of Applied Statistics
• /
• v.35 no.4
• /
• pp.469-484
• /
• 2022

Since a large text corpus contains hundred-thousand unique words, text data is one of the typical large-dimensional data. Therefore, various feature selection methods have been proposed for dimension reduction. Feature selection methods can improve the prediction accuracy. In addition, with reduced data size, computational efficiency also can be achieved. The chi-square statistic and the information gain are two of the most popular measures for identifying interesting terms from text data. In this paper, we investigate the theoretical properties of the chi-square statistic and the information gain. We show that the two filtering metrics share theoretical properties such as non-negativity and convexity. However, they are different from each other in the sense that the information gain is prone to select more negative features than the chi-square statistic in imbalanced text data.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.04b
• /
• pp.124-126
• /
• 2001

최근 데이터마이닝 응용분야에서 대용량의 고차원 데이터가 증가하고 있기 때문에 이를 효율적으로 처리할 수 있는 방법이 요구된다. 이를 위해 CLIQUE 방법과 셀-기반 클러스터링 방법을 선택하기 위해, 셀-기반 클러스터링 방법을 CLIQUE 방법 및 CLIQUE 방법에 근사정보(Approximation)를 결합한 방법과 성능 비교를 수행한다. 성능비교 결과, 셀-기반 클러스터링 방법이 데이터 클러스터링 및 데이터 검색시간에서 가장 우수한 성능을 보이며, 정확율은 CLIQUE 방법에 비해 다소 뒤떨어지거나 전체적인 효율성에서 매우 우수한 성능을 보인다.

• The KIPS Transactions:PartD
• /
• v.19D no.2
• /
• pp.151-160
• /
• 2012

Most of the distributed high-dimensional indexing structures provide a reasonable search performance especially when the dataset is uniformly distributed. However, in case when the dataset is clustered or skewed, the search performances gradually degrade as compared with the uniformly distributed dataset. We propose a method of improving the k-nearest neighbor search performance for the distributed vector approximation-tree based on the strongly clustered or skewed dataset. The basic idea is to compute volumes of the leaf nodes on the top-tree of a distributed vector approximation-tree and to assign different number of bits to them in order to assure an identification performance of vector approximation. In other words, it can be done by assigning more bits to the high-density clusters. We conducted experiments to compare the search performance with the distributed hybrid spill-tree and distributed vector approximation-tree by using the synthetic and real data sets. The experimental results show that our proposed scheme provides consistent results with significant performance improvements of the distributed vector approximation-tree for strongly clustered or skewed datasets.

• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.726-730
• /
• 2009

The proliferation of the web and digital photography has drastically increased multimedia data and has resulted in the need of the high quality internet service based on the moving picture like user generated contents(UGC). The keyword-based search on large scale images and video collections is too expensive and requires much manual intervention. Therefore the web search engine may provide the content-based retrieval on the multimedia data for search accuracy and customer satisfaction. In this paper, we propose a novel distributed index structure based on multiple length signature files according to data distribution. In addition, we describe how our scalable index technique can be used to find the nearest neighbors in the cluster environments.

• Journal of Internet of Things and Convergence
• /
• v.9 no.6
• /
• pp.119-124
• /
• 2023

Deep learning has recently demonstrated conspicuous efficacy across diverse domains than traditional machine learning techniques, as the most popular approach for pattern recognition. The classification problems for tabular data, however, are remain for the area of traditional machine learning. This paper introduces a novel network module designed to tabular data into high-dimensional tensors. The module is integrated into conventional deep learning networks and subsequently applied to the classification of structured data. The proposed method undergoes training and validation on four datasets, culminating in an average accuracy of 90.22%. Notably, this performance surpasses that of the contemporary deep learning model, TabNet, by 2.55%p. The proposed approach acquires significance by virtue of its capacity to harness diverse network architectures, renowned for their superior performance in the domain of computer vision, for the analysis of tabular data.

• The Korean Journal of Applied Statistics
• /
• v.37 no.2
• /
• pp.157-176
• /
• 2024

Integrative analysis for better understanding of complex biological systems gains more attention. Observing subjects from various perspectives and conducting integrative analysis of those multiple datasets enables a deeper understanding of the subject. In this paper, we compared two methods that simultaneously consider two datasets gathered from the same objects, canonical correlation analysis (CCA) and co-inertia analysis (CIA). Since CCA cannot handle the case when the data exhibit high-dimensionality, two strategies were considered instead: Utilization of a ridge constant (CCA-ridge) and substitution of covariance matrices of each data to identity matrix and then applying penalized singular value decomposition (CCA-PMD). To illustrate CIA and CCA, both extensions of CCA and CIA were applied to NCI60 cell line data. It is shown that both methods yield biologically meaningful and significant results by identifying important genes that enhance our comprehension of the data. Their results shows some dissimilarities arisen from the different criteria used to measure the relationship between two sets of data in each method. Additionally, CIA exhibits variations dependent on the weight matrices employed.