• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.2037-2040
• /
• 2004

In this paper we introduce a new prototype based classifiers for overlapping data, where training pattern can be overlap on the feature space. The proposed classifier is based on the prototype from neural network classifier (NNC)[1] for overlap data. The method automatically chooses the initial center and two radiuses for each class. The center is used as a mean representative of training data for each class. The unclassified pattern is classified by measure distance from the class center. If the distance is in the lower (shorter radius) the unknown pattern has the high percentage of being in this class. If the distance is between the lower and upper (further radius), the pattern has the probability of being in this class or others. But if the distance is outside the upper, the pattern is not in this class. We borrow the words upper and lower from the rough set to represent the region of certainty [3]. The training algorithm to find number of cluster and their parameters (center, lower, upper) is presented. The clustering result is tested using patterns from Thai handwritten letter and the clustering result is very similar to human eyes clustering.

• Journal of IKEEE
• /
• v.12 no.4
• /
• pp.217-224
• /
• 2008

This paper provides a framework for semantic correspondence of heterogeneous databases using self- organizing map. It solves the problem of overlapping between different databases due to their different schemas. Clustering technique using self-organizing maps (SOM) is tested and evaluated to assess its performance when using different kinds of data. Preprocessing of database is performed prior to clustering using edit distance algorithm, principal component analysis (PCA), and normalization function to identify the features necessary for clustering.

• Journal of Intelligence and Information Systems
• /
• v.17 no.1
• /
• pp.111-125
• /
• 2011

Clustering is a process of grouping similar or relevant documents into a cluster and assigning a meaningful concept to the cluster. By this process, clustering facilitates fast and correct search for the relevant documents by narrowing down the range of searching only to the collection of documents belonging to related clusters. For effective clustering, techniques are required for identifying similar documents and grouping them into a cluster, and discovering a concept that is most relevant to the cluster. One of the problems often appearing in this context is the detection of a complex concept that overlaps with several simple concepts at the same hierarchical level. Previous clustering methods were unable to identify and represent a complex concept that belongs to several different clusters at the same level in the concept hierarchy, and also could not validate the semantic hierarchical relationship between a complex concept and each of simple concepts. In order to solve these problems, this paper proposes a new clustering method that identifies and represents complex concepts efficiently. We developed the Hierarchical Overlapping Clustering (HOC) algorithm that modified the traditional Agglomerative Hierarchical Clustering algorithm to allow overlapped clusters at the same level in the concept hierarchy. The HOC algorithm represents the clustering result not by a tree but by a lattice to detect complex concepts. We developed a system that employs the HOC algorithm to carry out the goal of complex concept detection. This system operates in three phases; 1) the preprocessing of documents, 2) the clustering using the HOC algorithm, and 3) the validation of semantic hierarchical relationships among the concepts in the lattice obtained as a result of clustering. The preprocessing phase represents the documents as x-y coordinate values in a 2-dimensional space by considering the weights of terms appearing in the documents. First, it goes through some refinement process by applying stopwords removal and stemming to extract index terms. Then, each index term is assigned a TF-IDF weight value and the x-y coordinate value for each document is determined by combining the TF-IDF values of the terms in it. The clustering phase uses the HOC algorithm in which the similarity between the documents is calculated by applying the Euclidean distance method. Initially, a cluster is generated for each document by grouping those documents that are closest to it. Then, the distance between any two clusters is measured, grouping the closest clusters as a new cluster. This process is repeated until the root cluster is generated. In the validation phase, the feature selection method is applied to validate the appropriateness of the cluster concepts built by the HOC algorithm to see if they have meaningful hierarchical relationships. Feature selection is a method of extracting key features from a document by identifying and assigning weight values to important and representative terms in the document. In order to correctly select key features, a method is needed to determine how each term contributes to the class of the document. Among several methods achieving this goal, this paper adopted the $x^2$�� statistics, which measures the dependency degree of a term t to a class c, and represents the relationship between t and c by a numerical value. To demonstrate the effectiveness of the HOC algorithm, a series of performance evaluation is carried out by using a well-known Reuter-21578 news collection. The result of performance evaluation showed that the HOC algorithm greatly contributes to detecting and producing complex concepts by generating the concept hierarchy in a lattice structure.

• Journal of Institute of Control, Robotics and Systems
• /
• v.12 no.10
• /
• pp.1044-1049
• /
• 2006

Multiple sequence alignment is a method to compare two or more DNA or protein sequences. Most of multiple sequence alignment tools rely on pairwise alignment and Smith-Waterman algorithm to generate an alignment hierarchy. Therefore, in the existing multiple alignment method as the number of sequences increases, the runtime increases exponentially. In order to remedy this problem, we adopted a parallel processing suffix tree algorithm that is able to search for common subsequences at one time without pairwise alignment. Also, the cross-matching subsequences triggering inexact-matching among the searched common subsequences might be produced. So, the cross-matching masking process was suggested in this paper. To identify the function of the clusters generated by suffix tree clustering, BLAST and CDD (Conserved Domain Database)search were combined with a clustering tool. Our clustering and annotating tool consists of constructing suffix tree, overlapping common subsequences, clustering gene sequences and annotating gene clusters by BLAST and CDD search. The system was successfully evaluated with 36 gene sequences in the pentose phosphate pathway, clustering 10 clusters, finding out representative common subsequences, and finally identifying functional domains by searching CDD database.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2008.05a
• /
• pp.890-893
• /
• 2008

자율성 및 이동성 갖는 네트워크 구조의 하나인 MANET(Mobile Ad-Hoc Networks)은 각 node들은 그 특성에 따라서 clustering service을 한다. node의 전송과정 중 path access에 대하여 중요성 또한 강조되고 있다. 일반적인 무선 네트워크 상에서의 node들은 clustering을 하게 되는데 그 과정에서 발생되는 여러 가지 문제점을 가지고 전송이 이루어진다. 모든 node들이 송, 수신상의 전송 범위(Beam forming area)가지고 있으며, 이러한 각 node들의 전송범위 내에 전송이 이루어지는 전통적인 전송기술 mechanism을 찾는다. 이러한 전송상황에서의 송신하는 node와 수신된 node간에 발생되고 있는 중복성의 문제점으로 즉, 상호적용에 의한 네트워크 duplicate(overlapping)이 크게 우려가 되고 있다. 이러한 전송상의 전송 범위 중첩, node간의 packet 간섭현상, packet의 중복수신 및 broadcasting의 storming현상이 나타난다. 따라서 본 논문에서는 상황정보의 속성을 이용한 계층적 상호 head node들의 접근된 위치와 연계되는 전송속도, 보존하고 있는 head node들의 에너지 source value, doppler효과를 통한 head node의 이동방향 등 분석한다. 분석된 방법으로 전송상의 계층적 path가 구성된 경험적 path 속성을 통한 네트워크 connectivity 신뢰성을 극대화 할 뿐만 아니라 네트워크의 전송 범위 duplicate을 사전에 줄일 수 있고 전송망의 최적화를 유지할 수 있는 기법의 하나인 상황정보를 이용한 ad hoc network의 ODDMRP(Ontology Doppler effect-based Dynamic Multicast Routing Protocol) clustering 기법을 제안한다.

• The Journal of Korea Robotics Society
• /
• v.17 no.3
• /
• pp.255-263
• /
• 2022

3D object detection pipelines often incorporate RGB-based object detection methods such as YOLO, which detects the object classes and bounding boxes from the RGB image. However, in complex environments where objects are heavily cluttered, bounding box approaches may show degraded performance due to the overlapping bounding boxes. Mask based methods such as Mask R-CNN can handle such situation better thanks to their detailed object masks, but they require much longer time for data preparation compared to bounding box-based approaches. In this paper, we present a 3D object recognition pipeline which uses either the YOLO or Mask R-CNN real-time object detection algorithm, K-nearest clustering algorithm, mask reduction algorithm and finally Principal Component Analysis (PCA) alg orithm to efficiently detect 3D poses of objects in a complex environment. Furthermore, we also present an improved YOLO based 3D object detection algorithm that uses a prioritized heightmap clustering algorithm to handle overlapping bounding boxes. The suggested algorithms have successfully been used at the Artificial-Intelligence Robot Challenge (ARC) 2021 competition with excellent results.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07a
• /
• pp.427-429
• /
• 2005

Energy efficiency is the most challenging issue in wireless sensor network to prolong the life time of the network, as the sensors has to be unattended. Cluster based communication can reduce the traffic on the network and gives the opportunity to other sensors for periodic sleep and thus save energy. Passive clustering (PC) can perform a significant role to minimize the network load as it is less computational and light weight. First declaration wins method of PC without any priority generates severe collision in the network and forms the clusters very dense with large amount of overlapping region. We have proposed several modifications for the existing passive clustering algorithm to prolong the life time of the network with better cluster formation.

• Asian Pacific Journal of Cancer Prevention
• /
• v.15 no.14
• /
• pp.5721-5727
• /
• 2014

Objectives: To evaluate the performance of clustering methods used in the prognostic assessment of categorical clinical data for hepatocellular carcinoma (HCC) patients in China, and establish a predictable prognostic nomogram for clinical decisions. Materials and Methods: A total of 332 newly diagnosed HCC patients treated with hepatic resection during 2006-2009 were enrolled. Patients were regularly followed up at outpatient clinics. Clustering methods including the Average linkage, k-modes, fuzzy k-modes, PAM, CLARA, protocluster, and ROCK were compared by Monte Carlo simulation, and the optimal method was applied to investigate the clustering pattern of the indices including platelet count, platelet/lymphocyte ratio (PLR) and serum aspartate aminotransferase activity/platelet count ratio index (APRI). Then the clustering variable, age group, tumor size, number of tumor and vascular invasion were studied in a multivariable Cox regression model. A prognostic nomogram was constructed for clinical decisions. Results: The ROCK was best in both the overlapping and non-overlapping cases performed to assess the prognostic value of platelet-based indices. Patients with categorical platelet-based indices significantly split across two clusters, and those with high values, had a high risk of HCC recurrence (hazard ratio [HR] 1.42, 95% CI 1.09-1.86; p<0.01). Tumor size, number of tumor and blood vessel invasion were also associated with high risk of HCC recurrence (all p< 0.01). The nomogram well predicted HCC patient survival at 3 and 5 years. Conclusions: A cluster of platelet-based indices combined with other clinical covariates could be used for prognosis evaluation in HCC.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10a
• /
• pp.175-177
• /
• 2001

기존의 문서 클러스터링 알고리즘은 모든 문서가 각각 하나의 클러스터에만 할당되도록 설계되어 문서에 여러 개의 주제가 포함되어 있을지라도 문서는 유사도 비교에 의해 오직 하나의 플러스터에 포함된다는 단점이 있다. 본 연구에서는 이러한 문서 플러스터링 방법의 한계를 파악하기 위해 문서가 여러 개의 클러스터에 포함될 수 있는 계층적 중복 문서 클러스터링을 제안한다. 또한, 문서 클러스터링의 정확도를 높이기 위해서 불용어 제거 알고리즘을 이용해 불용어를 제거하여 클러스터링에 사용되는 키워드를 선별하고, 단어가중치 산출을 위한 TF*NHDF 공식을 제안한다.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.37 no.3
• /
• pp.259-266
• /
• 2013

An automated cell tracking system is used to automatically analyze and track the changes in cell behavior in time-lapse cell images acquired using a microscope with a cell culture. Clustering is the partial overlapping of neighboring cells in the process of cell change. Separating clusters into individual cells is very important for cell tracking. In this study, we proposed an algorithm for separating clusters by using ellipse fitting based on a direct least square method. We extracted the contours of clusters, divided them into line segments, and then produced their fitted ellipses using a direct least square method for each line segment. All of the fitted ellipses could be used to separate their corresponding clusters. In experiments, our algorithm separated clusters with average precisions of 91% for two overlapping cells, 84% for three overlapping cells, and about 73% for four overlapping cells.