• Journal of Information Science Theory and Practice
• /
• v.8 no.2
• /
• pp.6-17
• /
• 2020

Word similarity is often measured to enhance system performance in the information retrieval field and other related areas. This paper reports on an experimental comparison of values for word similarity measures that were computed based on 50 intentionally selected words from a Reuters corpus. There were three targets, including (1) co-occurrence-based similarity measures (for which a co-occurrence frequency is counted as the number of documents or sentences), (2) context-based distributional similarity measures obtained from a latent Dirichlet allocation (LDA), nonnegative matrix factorization (NMF), and Word2Vec algorithm, and (3) similarity measures computed from the tf-idf weights of each word according to a vector space model (VSM). Here, a Pearson correlation coefficient for a pair of VSM-based similarity measures and co-occurrence-based similarity measures according to the number of documents was highest. Group-average agglomerative hierarchical clustering was also applied to similarity matrices computed by individual measures. An evaluation of the cluster sets according to an answer set revealed that VSM- and LDA-based similarity measures performed best.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2009.04a
• /
• pp.399-400
• /
• 2009

As a part of the era of human centric applications people started to care about their well being utilizing any possible mean. This paper proposes a framework for real time on-body sensor health-care system, addresses the current issues in such systems, and utilizes an enhanced online divisive agglomerative clustering algorithm (EODAC); an algorithm that builds a top-down tree-like structure of clusters that evolves with streaming data to rationally cluster on-body sensor data and give accurate diagnoses remotely, guaranteeing high performance, and scalability. Furthermore it does not depend on the number of data points.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.95-96
• /
• 2021

최근 미디어의 발전으로 빠른 속도로 많은 양의 사람들의 얼굴이 포함된 사진, 동영상들이 인터넷에 업로드 되고 있다. 이러한 현상에 맞춰 인공지능을 활용한 얼굴 인식 기술의 놀라운 발전이 있었으나, 대규모 데이터셋에서 임의의 인물을 검색하는 경우에서는 연산량과 저장공간의 부담이 존재한다. 특히, 인터넷에 존재하는 수많은 불법 촬영물에서 피해자를 정확하고 신속하게 검색하기 위해서는 효율적인 얼굴 검색 시스템이 필요하다. 따라서, 본 논문은 얼굴 특징 추출과 클러스터링을 활용하여 방대한 양의 불법 촬영물 셋에서 피해자 동영상을 효율적으로 검색할 수 있는 기법을 제안한다. 불법 촬영물 동영상 검색 실험 환경을 만들기 위해 YouTube Faces [1] 데이터셋으로 유사 동영상 셋을 만들고 이 환경에서 실험을 진행한다. 얼굴 특징 추출 모델은 ResNet100 네트워크를 CosFace 손실함수와 Glint360K 데이터셋으로 학습시킨 모델 [2]을 사용한다. 추출된 얼굴 특징들을 HAC(Hierarchical Agglomerative Clustering) 알고리즘으로 클러스터링 한 후, 클러스터 대푯값을 통해 얼굴 검색 실험을 했을 때의 실험 결과를 분석한다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.8
• /
• pp.1155-1160
• /
• 2013

The rapid growth of smart phones has made location-based services (LBSs) widely available. However, the use of LBS can raise privacy issues, as LBS can allow adversaries to violate the location privacy of users. There has been a considerable amount of research on preserving user location privacy. Most of these studies try to preserve location privacy by achieving what is known as location K-anonymity. In this paper, we propose a hierarchical clustering-based spatial cloaking algorithm for LBSs. The proposed algorithm constructs a tree using a modified version of agglomerative hierarchical clustering. The experimental results show, in terms of the ASR size, that the proposed algorithm is better than Hilbert Cloak and comparable to RC-AR (R-tree Cloak implementation of Reciprocal with an Asymmetric R-tree split). In terms of the ASR generation time, the proposed algorithm is much better in its performance than RC-AR and similar in performance to Hilbert Cloak.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.10
• /
• pp.705-712
• /
• 2013

Active learning aims to improve the performance of a classification model by repeating the process to select the most helpful unlabeled data and include it to the training set through labelling by expert. In this paper, we propose a method for active learning based on hierarchical agglomerative clustering using Ward's linkage. The proposed method is able to construct a training set actively so as to include at least one sample from each cluster and also to reflect the total data distribution by expanding the existing training set. While most of existing active learning methods assume that an initial training set is given, the proposed method is applicable in both cases when an initial training data is given or not given. Experimental results show the superiority of the proposed method.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.7
• /
• pp.1733-1748
• /
• 1997

In this paper, we propose the classification scheme and retrieval mechanism which can apply to many application domains in order to construct the software reuse library. Classification scheme which is the core of the accessibility in the reusability, is defined by the hierarchical structure using the agglomerative clusters. Agglomerative cluster means the group of the reuse component by the functional relationships. Functional relationships are measured by the HACM which is the representation method about software components to calculate the similarities among the classes in the particular domain. And clustering informations are added to the library structure which determines the functionality and accuracy of the retrieval system. And the system stores the classification results such as the index information with the weights, the similarity matrix, the hierarchical structure. Therefore users can retrieve the software component using the query which is the natural language. The thesis is studied to focus on the findability of software components in the reuse library. As a result, the part of the construction process of the reuse library was automated, and we can construct the object-oriented reuse library with the extendibility and relationship about the reuse components. Also the our process is visualized through the browse hierarchy of the retrieval environment, and the retrieval system is integrated to the reuse system CARS 2.1.

• Journal of the Korean Statistical Society
• /
• v.20 no.2
• /
• pp.162-176
• /
• 1991

The problem of determining the number of clusters, K. is the main objective of this study. Attention is focused on the use of Rand(1971)'s $C_{k}$ statistic with some agglomerative clustering algorithms(ACA) defined in the ($\beta$, $\pi$) plane in predicting the number of clusters within the given set of data. The (k, $C_{k}$) plots for k=1, 2, …, N are explored by a Monte Carlo study. Based on its performance, the use of $C_{k}$ with the pair of ACA, (-.5, .75) and (-.25, .0), is recommended for predicting the number of clusters present within a set of data. data.

• Journal of Multimedia Information System
• /
• v.7 no.3
• /
• pp.197-204
• /
• 2020

Superpixel methods are widely used in the preprocessing stage as a method to reduce computational complexity by simplifying images while maintaining the characteristics of the images in the computer vision applications. It is common to generate superpixels of similar size and shape based on the pixel values rather than considering the characteristics of the image. In this paper, we propose a method to control the sizes and shapes of generated superpixels, considering the contents of an image. The proposed method consists of two steps. The first step is to over-segment an image so that the boundary information of the image is well preserved. In the second step, generated superpixels are merged based on similarity to produce the target number of superpixels, where the shapes of superpixels are controlled by limiting the maximum size and the proposed roundness metric. Experimental results show that the proposed method preserves the boundaries of the objects in an image more accurately than the existing method.

• The Korean Journal of Applied Statistics
• /
• v.18 no.2
• /
• pp.329-341
• /
• 2005

A method of determining the number of clusters is proposed based on some asymptotic results on the Rand's(1971} $C_k$, k = 2, 3, . . ., N - 1, statistic. Simulation is conducted to compare the proposed method with Chae and Warde(1991), and Huh and Lee(2004).

• Proceedings of the Korean Society for Information Management Conference
• /
• 2005.08a
• /
• pp.65-70
• /
• 2005

계층적 문서 클러스터링에 있어서 실험집단에 따라 응집식 기법과 분할식 기법의 성능이 다르며, 이를 좌우하는 요소는 분류의 깊이, 즉 분류수준이라고 가정하였다. 조금만 나누면 되는 대분류인 경우는 상대적으로 분할식 기법이 유리하고, 조금만 합치면 되는 소분류인 경우에는 응집식 기법이 유리할 것이라고 판단했기 때문이다. 그에 따라 분할식 클러스터링 기법인 양분(Bisecting) K-means기법과 응집식 기법인 완전연결, 평균연결, WARD기법의 성능을 실험집단이 대분류인 경우와 소분류인 경우의 유사계수를 적용하여 각 기법별 성능을 비교하여 실험집단의 특성에 따른 적합 클러스터링 기법을 찾고자 하였다. 실험결과 응집식 기법과 분할식 기법의 성능 우열에 영향을 미치는 것은 분류수준보다는 변이계수로 측정된 상대적인 군집의 크기 편차인 것으로 나타났다.