Search

Current approaches to improve the pattern recognition performance mainly focus on either extracting non-stationary and discriminant features of each class, or employing complex and nonlinear feature classifiers. However, little attention has been paid to the integration of these two approaches. Combining non-stationary feature analysis with complex feature classifiers, this article presents a novel direction to enhance the discriminatory power of pattern recognition methods. This approach, which is based on a fusion of non-stationary feature analysis with clustering techniques, proposes an algorithm to adaptively identify the feature vectors according to their importance in representing the patterns of discrimination. Non-stationary feature vectors are extracted using a non-stationary method based on time–frequency distribution and non-negative matrix factorization. The clustering algorithms including the K-means and self-organizing tree maps are utilized as unsupervised clustering methods followed by a supervised labeling. Two labeling methods are introduced: hard and fuzzy labeling. The article covers in detail the formulation of the proposed discriminant feature clustering method. Experiments performed with pathological speech classification, T-wave alternans evaluation from the surface electrocardiogram, audio scene analysis, and telemonitoring of Parkinson’s disease problems produced desirable results. The outcome demonstrates the benefits of non-stationary feature fusion with clustering methods for complex data analysis where existing approaches do not exhibit a high performance.

In this work, we will discuss how to improve Wikilabel, an approach which makes use of titles in Wikipedia pages to generate labels for documents, by retooling ideas from story link detection (SLD). A comparison of our approach against Elastic Net, a powerful machine learner, on the real world data, finds the visible superiority of our approach over the latter.

We propose a competence partitioning strategy for Web search result presentation: the unmodified head of a ranked result list is combined with a clustering of documents from the result list tail. We identify two principles to which such a clustering must adhere to improve the user's search experience: (1) Avoid the unwanted effect of query aspect repetition, which is called shadowing here. (2) Avoid extreme clusterings, i.e., neither the number of cluster labels nor the number of documents per cluster should exceed the size of the result list head. We present measures to quantify the shadowing effect, and with Faceted Clustering we introduce an algorithm that optimizes the identified principles. The key idea of Faceted Clustering is a dynamic, user-controlled reorganization of a clustering, similar to a faceted navigation system. We report on evaluations using the AMBIENT corpus and demonstrate the potential of our approach by a comparison with two well-known clustering search engines.

El clustering de documentos es un campo de investigación popular en los ámbitos del Procesamiento del Lenguaje Natural, la Minería de Datos y la Recuperación de información (RI). El problema de agrupar unidades léxicas mediante clustering ha sido menos estudiado y menos aún, el problema de etiquetar los clusters. Sin embargo, en nuestra aplicación que trata sobre la extracción de tuplas de relaciones para ser usadas como entrada a programas para dibujar diagramas de bloques o mapas conceptuales, este problema es fundamental. La valoración de varias estrategias de etiquetado de clusters de documentos nos revela que algunas de estas técnicas pueden ser también aplicadas para etiquetar nuestros clusters, compuestos por verbos semánticamente similares. Para confirmar esta suposición, llevamos a cabo una serie de experimentos y evaluamos su rendimiento contra baselines y un goldstandard de clusters etiquetados. Document clustering is a popular research field in Natural Language Processing, Data Mining and Information Retrieval. The problem of lexical unit (LU) clustering has been less addressed, and even less so the problem of labeling LU clusters. However, in our application that deals with the distillation of relational tuples from patent claims as input to block diagram or a concept map drawing programs, this problem is central. The assessment of various document cluster labeling techniques lets us assume that despite some significant differences that need to be taken into account some of these techniques may also be applied to verbal relation cluster labeling we are concerned with. To confirm this assumption, we carry out a number of experiments and evaluate their outcome against baselines and gold standard labeled clusters.

Clustering is of great practical value in retrieving reusable requirements artifacts from the ever-growing software project repositories. Despite the development of automated cluster labeling techniques in information retrieval, little is understood about automatic labeling of requirements clusters. In this paper, we review the literature on cluster labeling, and conduct an experiment to evaluate how automated methods perform in labeling requirements clusters. The results show that differential labeling outperforms cluster-internal labeling, and that hybrid method does not necessarily lead to the labels best matching human judgment. Our work sheds light on improving automated ways to support search-driven development.

The paper addresses the missing user acceptance of web search result clustering. We report on selected analyses and propose new concepts to improve existing result clustering approaches. Our findings in a nutshell are: 1. Don't compete with a search engine's top hits. In response to a query we presume search engines to return an optimal result list in the sense of the probabilistic ranking principle: documents that are expected by the majority of users are placed on top and form the result list head. We argue that, with respect to the top results, it is not beneficial to replace this established form of result presentation. 2. Improve document access in the result list tail. Documents that address the information need of "minorities" appear at some position in the result list tail. Especially for ambiguous and multi-faceted queries we expect this tail to be long, with many users appreciating different documents. In this situation web search result clustering can improve user satisfaction by reorganizing the long tail into topic-specific clusters. 3. Avoid shadowing when constructing cluster labels. We show that most of the cluster labels that are generated by current clustering technology occur within the snippets of the result list head--an effect which we call shadowing. The value of such labels for topic organization and navigating within a clustering of the entire result list is limited. We propose and analyze a filtering approach to significantly alleviate the label shadowing effect.

This paper presents a particular approach to collective labeling of multiple documents, which works by associating the documents with Wikipedia pages and labeling them with headings the pages carry. The approach has an obvious advantage over past approaches in that it is able to produce fluent labels, as they are hand-written by human editors. We carried out some experiments on the TDT5 dataset, which found that the approach works rather robustly for an arbitrary set of documents in the news domain. Comparisons were made with some baselines, including the state of the art, with results strongly in favor of our approach.

A simple and intuitive way to organize a huge document collection is by a topic hierarchy. Generally two steps are carried out to build a topic hierarchy automatically: 1) hierarchical document clustering and 2) cluster labeling. For both steps, a good textual document representation is essential. The bag-of-words is the common way to represent text collections. In this representation, each document is represented by a vector where each word in the document collection represents a dimension (feature). This approach has well known problems as the high dimensionality and sparsity of data. Besides, most of the concepts are composed by more than one word, as "document engineering" or "text mining". In this paper an approach called bag-of-related-words is proposed to generate features compounded by a set of related words with a dimensionality smaller than the bag-of-words. The features are extracted from each textual document of a collection using association rules. Different ways to map the document into transactions in order to allow the extraction of association rules and interest measures to prune the number of features are analyzed. To evaluate how much the proposed approach can aid the topic hierarchy building, we carried out an objective evaluation for the clustering structure, and a subjective evaluation for topic hierarchies. All the results were compared with the bag-of-words. The obtained results demonstrated that the proposed representation is better than the bag-of-words for the topic hierarchy building.

International audience; Neural clustering algorithms show high performance in the general context of the analysis of homogeneous textual dataset. This is especially true for the recent adaptive versions of these algorithms, like the incremental growing neural gas algorithm (IGNG) and the labeling maximization based incremental growing neural gas algorithm (IGNG-F). In this paper we highlight that there is a drastic decrease of performance of these algorithms, as well as the one of more classical algorithms, when a heterogeneous textual dataset is considered as an input. Specific quality measures and cluster labeling techniques that are independent of the clustering method are used for the precise performance evaluation. We provide new variations to incremental growing neural gas algorithm exploiting in an incremental way knowledge from clusters about their current labeling along with cluster distance measure data. This solution leads to significant gain in performance for all types of datasets, especially for the clustering of complex heterogeneous textual data.

Cluster identification tasks occur in a multitude of contexts in physics and engineering such as, for instance, cluster algorithms for simulating spin models, percolation simulations, segmentation problems in image processing, or network analysis. While it has been shown that graphics processing units (GPUs) can result in speedups of two to three orders of magnitude as compared to serial codes on CPUs for the case of local and thus naturally parallelized problems such as single-spin flip update simulations of spin models, the situation is considerably more complicated for the nonlocal problem of cluster or connected component identification. I discuss the suitability of different approaches of parallelization of cluster labeling and cluster update algorithms for calculations on GPU and compare to the performance of serial implementations.

Neural clustering algorithms show high performance in the general context of the analysis of homogeneous textual dataset. This is especially true for the recent adaptive versions of these algorithms, like the incremental growing neural gas algorithm (IGNG) and the label maximization based incremental growing neural gas algorithm (IGNG-F). In this paper we highlight that there is a drastic decrease of performance of these algorithms, as well as the one of more classical algorithms, when a heterogeneous textual dataset is considered as an input. Specific quality measures and cluster labeling techniques that are independent of the clustering method are used for the precise performance evaluation. We provide variations to incremental growing neural gas algorithm exploiting in an incremental way knowledge from clusters about their current labeling along with cluster distance measure data. This solution leads to significant gain in performance for all types of datasets, especially for the clustering of complex heterogeneous textual data.

Ankara : The Department of Computer Engineering and the Graduate School of Engineering and Science of Bilkent University, 2011. Thesis (Master's) -- Bilkent University, 2011. Includes bibliographical references leaves 58-62. Search engines present query results as a long ordered list of web snippets divided into several pages. Post-processing of information retrieval results for easier access to the desired information is an important research problem. A post-processing technique is clustering search results by topics and labeling these groups to reflect the topic of each cluster. In this thesis, we present a novel search result clustering approach to split the long list of documents returned by search engines into meaningfully grouped and labeled clusters. Our method emphasizes clustering quality by using cover coefficient and sequential k-means clustering algorithms. Cluster labeling is crucial because meaningless or confusing labels may mislead users to check wrong clusters for the query and lose extra time. Additionally, labels should reflect the contents of documents within the cluster accurately. To be able to label clusters effectively, a new cluster labeling method based on term weighting is introduced. We also present a new metric that employs precision and recall to assess the success of cluster labeling. We adopt a comparative evaluation strategy to derive the relative performance of the proposed method with respect to the two prominent search result clustering methods: Suffix Tree Clustering and Lingo. Moreover, we perform the experiments using the publicly available Ambient and ODP-239 datasets. Experimental results show that the proposed method can successfully achieve both clustering and labeling tasks. Türel, Anıl M.S.

Cluster algorithms are non-local Monte Carlo update schemes which can greatly increase the efficiency of computer simulations of spin models of magnets. The major computational task in these algorithms is connected component labeling, to identify clusters of connected sites on a lattice. We have devised some new SIMD component labeling algorithms, and implemented them on the Connection Machine. We investigate their performance when applied to the cluster update of the two-dimensional Ising spin model. These algorithms could also be applied to other problems which use connected component labeling, such as percolation and image analysis.

Targeting the same objective of alleviating the manual work as automatic annotation, in this paper, we propose a novel framework with minimal human effort to manually annotate a large-scale image corpus. In this framework, a dynamic multi-scale cluster labeling strategy is proposed to manually label the clusters of similar image regions. The users label the multi-scale clusters of regions instead of individual images, thus each labeling operation can annotate hundreds or even thousands of images simultaneously with much reduced manual work. Meanwhile the manual labeling guarantees the accuracy of the labels. Compared to automatic annotation, the proposed framework is more flexible, general and effective, especially for annotating those labels with large semantic gaps. Experiments on NUS-WIDE dataset demonstrate that the proposed fast manual annotation framework is much more effective than automatic annotation and comparatively efficient.

This paper compares four commonly used fuzzy analytical methods for remote sensing digital image classification, i.e. fuzzy c-means, semi-supervised fuzzy cluster labeling, fuzzy nearest neighbor, and object-oriented fuzzy classifiers. Merits and weak points of each method were examined through a case study with a multispectral high-resolution airborne digital image of urban settings. Results showed that the fuzzy labeling approach produced the highest quality, which was followed by the object-oriented fuzzy classifier. As the former combines merits of supervised and unsupervised classifications, the latter takes the full account of contextual and spatial features.

A multiple-perspective co-citation analysis method is introduced for characterizing and interpreting the structure and dynamics of co-citation clusters. The method facilitates analytic and sense making tasks by integrating network visualization, spectral clustering, automatic cluster labeling, and text summarization. Co-citation networks are decomposed into co-citation clusters. The interpretation of these clusters is augmented by automatic cluster labeling and summarization. The method focuses on the interrelations between a co-citation cluster's members and their citers. The generic method is applied to a three-part analysis of the field of Information Science as defined by 12 journals published between 1996 and 2008: 1) a comparative author co-citation analysis (ACA), 2) a progressive ACA of a time series of co-citation networks, and 3) a progressive document co-citation analysis (DCA). Results show that the multiple-perspective method increases the interpretability and accountability of both ACA and DCA networks., Comment: 33 pages, 11 figures, 10 tables. To appear in the Journal of the American Society for Information Science and Technology

Document clustering is a powerful technique that has been widely used for organizing data into smaller and manageable information kernels. Several approaches have been proposed suffering however from problems like synonymy, ambiguity and lack of a descriptive content marking of the generated clusters. We are proposing the enhancement of standard kmeans algorithm using the external knowledge from WordNet hypernyms in a twofold manner: enriching the "bag of words" used prior to the clustering process and assisting the label generation procedure following it. Our experimentation revealed a significant improvement over standard kmeans for a corpus of news articles derived from major news portals. Moreover, the cluster labeling process generates useful and of high quality cluster tags.

Digital libraries allow organizing, classifying and publishing collections of electronic contents that are available in computers or networks. Also, digital libraries are easy to use and configure and they offer a user interface with access to fast searching and browsing over a repository of documents using a graphical interface. This article presents a digital library prototype for retrieving, indexing and clustering documents published on a website. The website may include unstructured, semi-structured and structured documents such as: web pages, scientific papers, news and documents in several formats that contain essentially text. The proposed prototype includes a clustering process that uses a conceptual algorithm and an a priori process of cluster labeling. Preliminary results correspond to tests made with different sets of documents published in a real website.

We present an algorithm for cluster dynamics to efficiently simulate large systems on MIMD parallel computers with large numbers of processors. The method divides physical space into rectangular cells which are assigned to processors and combines a serial local procedure with a nearest neighbor relaxation process. By controlling overhead and reducing inter-processor communication this method attains good performance and speed-up. The complexity and scaling properties of the algorithm are analyzed. The algorithm has been used to simulate large 2d Ising systems (up to 27808 X 27808 sites) with Swendsen-Wang dynamics. Typical updating times on the order of 82 nanosecs/site and efficiencies larger than 90% have been obtained using 256 processors on the CM-5 supercomputer., 16 pages (TeX macros included)

We introduce a simple multi-scale algorithm for connected component labeling on parallel computers, which we apply to the problem of labeling clusters in spin and percolation models. We give numerical evidence that it is only logarithmically slowed down in the critical limit. We also discuss, in light of the proposed Teraflop computers optimized for lattice gauge theories and other lattice problems, the minimum requirements for computer switchboard architectures for which one can efficiently implement such multi-scale algorithms.

This work investigates cluster labeling enhancement by utilizing Wikipedia, the free on-line encyclopedia. We describe a general framework for cluster labeling that extracts candidate labels from Wikipedia in addition to important terms that are extracted directly from the text. The "labeling quality" of each candidate is then evaluated by several independent judges and the top evaluated candidates are recommended for labeling. Our experimental results reveal that the Wikipedia labels agree with manual labels associated by humans to a cluster, much more than with significant terms that are extracted directly from the text. We show that in most cases even when human's associated label appears in the text, pure statistical methods have difficulty in identifying them as good descriptors. Furthermore, our experiments show that for more than 85% of the clusters in our test collection, the manual label (or an inflection, or a synonym of it) appears in the top five labels recommended by our system.

Results clustering in Web Searching is useful for providing users with overviews of the results and thus allowing them to restrict their focus to the desired parts. However, the task of deriving single-word or multiple-word names for the clusters (usually referred as cluster labeling ) is difficult, because they have to be syntactically correct and predictive. Moreover efficiency is an important requirement since results clustering is an online task. Suffix Tree Clustering (STC) is a clustering technique where search results (mainly snippets) can be clustered fast (in linear time), incrementally, and each cluster is labeled with a phrase. In this paper we introduce: (a) a variation of the STC, called STC+, with a scoring formula that favors phrases that occur in document titles and differs in the way base clusters are merged, and (b) a novel non merging algorithm called NM-STC that results in hierarchically organized clusters. The comparative user evaluation showed that both STC+ and NM-STC are significantly more preferred than STC, and that NM-STC is about two times faster than STC and STC+.

Support vector clustering (SVC) is an appealing approach that can detect cluster boundaries. In spite of its popularization in applications, it sees the critical bottleneck in cluster labeling. This paper presents a Novel Support Vector Clustering algorithm (NSVC) to go a further step in clustering labeling. NSVC consists of three phases: extract Data Representatives (DRs); cluster DRs; label non-DR data. The objective of traditional SVC is used by NSVC for finding DRs, but the Kernel scale of the objective is modified. DRs are grouped by Spectrum Analysis (SA) method, which simultaneously develops an informative metric. Non-DR data are labeled by a weighted kNN procedure that works in query’s neighborhood, which is formulated with the new metric and then enriched by the convex hull skill. Experiments on real datasets demonstrate the improvement of NSVC over its peers and the competitive performance with the state of the arts.

A new parallel cluster-finding algorithm is formulated by using multigrid relaxation methods very similar to those used for differential equation solvers. For percolation clusters, this approach drastically reduces critical slowing down relative to local or scan relaxation methods. Numerical studies of scaling properties with system size are presented in the case of the 2D percolation clusters of the Swendsen-Wang Ising dynamics running on the Connection Machine.

In this chapter, the applications of rough set theory (RST) in text mining are discussed and a new concept named “Rough Text” is presented along with some RST-based measures for the evaluation of decision systems. We will focus on the application of such concept in clustering validity, specifically cluster labeling and multi-document summarization. The experimental studies show that the proposed measures outperform several internal measures existing on literature. Additionally, the application of Rough Text is illustrated.

In this paper we present a novel approach for labeling clusters of multimedia content that leverages supervised classification techniques in conjunction with unsupervised clustering. Recent research has produced significant results for automatic tagging of video content such as broadcast news. For example, powerful techniques have been demonstrated in the context of the NIST TRECVID video retrieval benchmark [1]. However, the information needs of users typically span a range of semantic concepts. One of the challenges of these multimedia retrieval systems is to organize the video data in such a way that allows the user to most efficiently navigate the semantic space for the video data set. One important tool for video data organization is clustering. However, clustering results cannot be leveraged effectively when they are not labeled. We propose to build on clustering by aggregating the automatically tagged semantics. We propose and compare four techniques for labeling the clusters and evaluate the performance compared to human labeled ground-truth. We present examples of the cluster labeling results obtained on the BBC stock shots from the TRECVID-2005 video data set.

In this demo we present a novel approach for labeling clusters in minimally annotated data archives. We propose to build on clustering by aggregating the automatically tagged semantics. We propose and compare four techniques for labeling the clusters and evaluate the performance compared to human labeled ground-truth. We define the error measures to quantify the results, and present examples of the cluster labeling results obtained on the BBC stock shots and broadcast news videos from the TRECVID-2005 video data set.

We describe ASK-GraphView, a node-link-based graph visualization system that allows clustering and interactive navigation of large graphs, ranging in size up to 16 million edges. The system uses a scalable architecture and a series of increasingly sophisticated clustering algorithms to construct a hierarchy on an arbitrary, weighted undirected input graph. By lowering the interactivity requirements we can scale to substantially bigger graphs. The user is allowed to navigate this hierarchy in a top down manner by interactively expanding individual clusters. ASK-GraphView also provides facilities for filtering and coloring, annotation and cluster labeling.

The support vector clustering (SVC) algorithm was introduced to get the number of the pinnacles in the result of the time-frequency analysis and Radon transform of the multi-component linear FM (LFM) signal, and to finish the detection of the components of the LFM signal. Meanwhile, the preprocessing to reduce the points' number of the input data-set for SVC, was proposed to improve the computation efficiency. And a novel cluster labeling method was developed to improve the SVC algorithm. The simulation results depict that the SVC-Radon-time-frequency approach is efficient for the detection and parameter estimation of the multi-component LFM signal with low SNR

Government agencies must often quickly organize and analyze large amounts of textual information, for example comments received as part of notice and comment rulemaking. Hierarchical organization is popular because it represents information at different levels of detail and is convenient for interactive browsing. Good hierarchical clustering algorithms are available, but there are few good solutions for automatically labeling the nodes in a cluster hierarchy.This paper presents a simple algorithm that automatically assigns labels to hierarchical clusters. The algorithm evaluates candidate labels using information from the cluster, the parent cluster, and corpus statistics. A trainable threshold enables the algorithm to assign just a few high-quality labels to each cluster. Experiments with Open Directory Project (ODP) hierarchies indicate that the algorithm creates cluster labels that are similar to labels created by ODP editors.

A cluster labeling algorithm for creating generic titles based on external resources such as WordNet is proposed. Our method first extracts category-specific terms as cluster descriptors. These descriptors are then mapped to generic terms based on a hypernym search algorithm. The proposed method has been evaluated on a patent document collection and a subset of the Reuters-21578 collection. Experimental results revealed that our method performs as anticipated. Real-case applications of these generic terms show promising in assisting humans in interpreting the clustered topics. Our method is general enough such that it can be easily extended to use other hierarchical resources for adaptable label generation.

This paper describes Armil, a meta-search engine that groups the web snippets returned by auxiliary search engines into disjoint labeled clusters. The cluster labels generated by Armil provide the user with a compact guide to assessing the relevance of each cluster to his/her information need. Striking the right balance between running time and cluster well-formedness was a key point in the design of our system. Both the clustering and the labeling tasks are performed on the fly by processing only the snippets provided by the auxiliary search engines, and they use no external sources of knowledge. Clustering is performed by means of a fast version of the furthest-point-first algorithm for metric $k$-center clustering. Cluster labeling is achieved by combining intra-cluster and inter-cluster term extraction based on a variant of the information gain measure. We have tested the clustering effectiveness of Armil against Vivisimo, the de facto industrial standard in web snippet clustering, using as benchmark a comprehensive set of snippets obtained from the Open Directory Project hierarchy. According to two widely accepted ``external'' metrics of clustering quality, Armil achieves better performance levels by $10%$. We also report the results of a thorough user evaluation of both the clustering and the cluster labeling algorithms. On a standard desktop PC (AMD Athlon 1-Ghz Clock with 750 Mbytes RAM), Armil performs clustering and labeling altogether of up to 200 snippets in less than one second.

Software architecture refers to the overall structure of a software system, and is defined by the components (sub-systems) within a software system and their interactions with one another. Quite often, there is little documentation describing a software system's architecture, especially in the case of legacy software systems. Thus techniques must be employed for recovering the architecture from the software's source code. Given the size and complexity of legacy systems, researchers have started exploring the use of automated techniques for architecture recovery. A technique that has shown promising results is clustering. Clusters that are obtained as a result of the clustering process represent sub-systems within a software system, but are nor easy to interpret until they are given appropriate names. In this paper, we present a cluster labeling scheme based on identifiers. As the clustering process proceeds, keywords are ranked using the inverse document frequency ranking scheme. Results of experiments conducted on a test system demonstrate that our labeling approach is effective. We also compare the clustering results of the complete algorithm and the weighted combined algorithm based on labels of the clusters produced by them during clustering.

In this paper, we present Stepping Stones and Pathways (SSP), an alternative model of building and presenting answers for the cases when queries on document collections cannot be answered just by a ranked list. Stepping Stones can handle questions like: "What is the relation of topics X and Y?" SSP addresses when the contents of a small set of related documents is needed as an answer rather than a single document, or when "query splitting" is required to satisfactorily explore a document space. Query results are networks of document groups representing topics, each group relating to and connecting (by documents) to other groups in the network. Thus, a network answers the user's information need. We devise new and more effective representations and techniques to visualize such answers, and to involve users as part of the answer-finding process. In order to verify the validity of our approach, and since the questions we aim to answer involve multiple topics, we performed a study involving a custom built broad collection of operating systems research papers, and evaluated the results with interested computer science students, using multiple measures.

The Enhancement Classification Method (ECM) has demonstrated considerable success in mapping Canada's forests and here is extended to facilitate clustering and labeling within a two-date classification. Central to the method, is an interactive cluster formulation based upon color rendition. A multi-date image enhancement is employed to facilitate an RGB rendition of change and cluster sieving is undertaken through spatial generalization. The remaining core clusters are reapplied via a minimum spectral distance. The method was tested on a 1984 - 1988 co-registered and normalized Landsat scene pair over a forest harvesting area near Petawawa, Ontario. Clusters (123) were derived and labeled. Results identified forest depletion (clear cuts and partial cuts) fairly well and captured stable forest composition moderately well and pre-change cover type moderately well. Burns, hail, forest blowdown and deforestation events were not recognized individually in single clusters, rather they were generally lumped into clearing classes. Consequently depletion requires a composite of classes to establish an intensity and localized cluster labeling.

With the rapid increase of data in many areas, clustering on large datasets has become an important problem in data analysis. Since cluster analysis is a highly iterative process, cluster analysis on large datasets prefers short iteration on a relatively small representative set. Thus, a two-phase framework "sampling/summarization - iterative cluster analysis" is often applied in practice. Since the clustering result only labels the small representative set, there are problems with extending the result to the entire large dataset, which are almost ignored by the traditional clustering research. This extending is often named as labeling process. Labeling irregular shaped clusters, distinguishing outliers and extending cluster boundary are the main problems in this stage. We address these problems and propose a visualization-based approach to dealing with them precisely. This approach partially involves human into the process of defining and refining the structure "ClusterMap". Based on this structure, the ClusterMap algorithm scans the large dataset to adapt the boundary extension and generate the cluster labels for the entire dataset. Experimental result shows that ClusterMap can preserve cluster quality considerably with low computational cost, compared to the distance-comparison-based labeling algorithms.

Interactive query refinement is widely believed to improve the effectiveness of ranked retrieval but it can be difficult to leverage existing batch evaluation frameworks to quantify the relative benefits of alternative interaction designs. This paper uses the new two-pass interaction design of the Text Retrieval Conference's High Accuracy Retrieval from Documents (HARD) track to explore the design space for cluster-based interactive relevance feedback. Two sites contributed two techniques for cluster formation and three techniques for cluster labeling. The effectiveness of each technique was compared with lower bounds based on blind relevance feedback, and with upper bounds found with oracle-based techniques. The clustering techniques were found to yield potential benefits, but the automatically constructed cluster labels were found not to support sufficiently accurate cluster selection. Elicitation of a desired cluster descriptor was found to significantly improve the effectiveness of a subsequent retrieval pass. These results indicate that the affordable two-pass study design used in the HARD track can yield useful insights to guide future design decisions.

Over the past decade, a number of initiatives have been undertaken to create definitive national and global data sets consisting of precision corrected Landsat Multispectral Scanner [MSS) and Thematic Mapper (TM) scenes. One important application of these data is the derivation of large area landcover products spanning multiple satellite scenes. A popular approach to land-cover mapping on this scale involves merging constituent scenes into image mosaics prior to image clustering and cluster labeling, thereby eliminating redundant geographic coverage arising from overlapping imaging swaths of adjacent orbital tracks. In this paper, arguments are presented to support the view that areas of overlapping coverage contain important information that can be used to assess and improve classification performance. A methodology is presented for the creation of large area land-cover products through the compositing of independently classified scenes. Statistical analyses of classification consistency between scenes in overlapping regions are employed both to identify mislabeled clusters and to provide a measure of classification confidence for each scene at the cluster level. During classification compositing, confidence measures are used to rationalize conflicting classifications in overlap regions and to create a relative confidence layer, sampled at the pixel level, which characterizes the spatial variation in classification quality over the final product. The procedure is illustrated with results from a synoptic mapping project of the Great Lakes watershed that involved the classification and compositing of 46 Landsat MSS scenes.

We present a rapid algorithm for identifying the current-carrying backbone in the percolation model. It applies to general two-dimensional graphs with open boundary conditions. Complemented by the modified Hoshen-Kopelman cluster labeling algorithm, our algorithm identifies dangling parts using their local properties. For planar graphs, it finds the backbone almost four times as fast as Tarjan's depth-first-search algorithm, and uses the memory of the same size as the modified Hoshen-Kopelman algorithm. Comparison with other algorithms for backbone identification is addressed., 5 pages with 5 eps figures. RevTeX 3.1. Clarify the origin of the hull-generating algorithm

Gait analysis is popular in many clinical and biomechanical applications, such as diagnosis of diabetic neuropathy, rehabilitation evaluation of stroke patients and performance measurement of sports training. With the rapid and in-depth development of flexible sensing technology, a large-area pressure-sensitive floor can be easily installed in many locations. Complex movement besides linear walking can be designed on large-area floors for gait analysis in clinical or sports research. To conduct those researches, as a basic step, a computational approach is necessary to track each footprint correctly during the movement process. A multi-stage methodology is proposed to solve two main subtasks in the tracking process: (1) the labeling of different footprints and (2) the detection of basic foot gestures in the movement process. The methodology consists of an initial clusters creating stage, a cluster labeling stage and an overlapped footprints separating process. Tai Chi Chuan, one of complex foot movements, was used as an example to evaluate the proposed approach. An overall accuracy of 99.07% for footprint labeling and 90.39% for basic foot gesture detecting were achieved by the method.

Publisher Summary The unique power of the scanning transmission electron microscopy (STEM) technique is its ability to examine isolated unstained structures in projection by mapping boundaries, internal mass distribution, and site-specific cluster labels, and measuring total mass. The digital STEM image can be compared directly to a computed projection of a model structure assembled from known or postulated components. If any feature of the proposed model is incorrect, the STEM image will provide direct statistical evidence as to the extent and significance of the discrepancy. STEM mass measurement, mass mapping, and heavy atom cluster labeling are widely known and accepted in the biophysics and biochemistry communities. The feasibility of the STEM portion of any proposed study provides the basic parameters of a possible model. It is similarly easy to assess how close the sample preparation is to the point of useful data collection with a simple pilot study. Therefore, the current approach is to try as many pilot studies of potentially feasible projects as possible and proceed as quickly as possible with those having the fewest practical problems.

We present an optimized version of a cluster labeling algorithm previously introduced by the authors. This algorithm is well suited for large-scale Monte Carlo simulations of spin models using cluster dynamics on parallel computers with large numbers of processors. The algorithm divides physical space into rectangular cells which are assigned to processors and combines a serial local labeling procedure with a relaxation process across nearest-neighbor processors. By controlling overhead and reducing inter-processor communication this method attains good computational speed-up and efficiency. Large systems of up to 65536 X 65536 spins have been simulated at updating speeds of 11 nanosecs/site (90.7 million spin updates/sec) using state-of-the-art supercomputers. In the second part of the article we use the cluster algorithm to study the relaxation of magnetization and energy on large Ising models using Swendsen-Wang dynamics. We found evidence that exponential and power law factors are present in the relaxation process as has been proposed by Hackl et al., 11 pages of LaTeX. Additional 17 pages of postscript figures are provided in a separated uuencoded file. A full Postcript version of this paper and other related papers are available at http://www.think.com/ProdServ/ADC/projects/proj.cc_spmd.html

This paper presents some approaches for the problem of labeling clusters using information from a given set of labeled and unlabeled patterns. Assigning class labels to the clusters is formulated as finding the best label assignment over all possible label assignments with respect to a criterion. Labeling clusters is also viewed as obtaining probabilities of class labels to the clusters with the maximization of the likelihood function and probability of correct labeling as criteria. Closed form solutions are obtained for the probabilities of class labels to the clusters by maximizing a lower bound on the likelihood criterion. Fixed point iteration equations are developed for obtaining probabilities of class labels to the clusters. The problem of obtaining class labels to the clusters is further formulated as that of minimizing the variance of the proportion estimates of the classes that use both the given labeled and unlabeled patterns. Imperfections in the labels of the given labeled set are incorporated into the criteria. Furthermore, the results of application of these techniques in the processing of remotely sensed multispectral scanner imagery data are presented.

In this paper the author considers the problem of obtaining the probabilities of class labels for the clusters using spectral and spatial information from a given set of labeled patterns and their neighbors. A relationship is developed between class and cluster conditional densities in terms of probabilities of class labels for the clusters. Expressions are presented for updating the a posteriori probabilities of the classes of a pixel using information from its local neighborhood. Fixed-point iteration schemes are developed for obtaining the optimal probabilities of class labels for the clusters. These schemes utilize spatial information and also the probabilities of label imperfections. Furthermore, experimental results from the processing of remotely sensed multispectral scanner imagery data are presented.