In model-driven engineering, metamodels may get lost over time resulting in the inability to load and view existing model instances. MARS is a system that recovers metamodels from model instances using grammar inference. This paper discusses advances in MARS that improve accuracy and scalability.

Content-based image retrieval has become an important part of information retrieval technology. Images can be viewed as high dimensional data and are usually represented by their low-level features. How to effectively find the semantic meanings of images is a central challenge in the area. In this paper, we propose an interactive platform for region-based image clustering and retrieval. A genetic algorithm is used to perform the initial clustering. In order to further refine the clustering results, we adopt the maximum flow/minimum cut theorem from graph theory to do outlier/outlier group detection. Outlier detection can help identify misclustered image segments and is used to improve the quality of clusters in this paper. In the interactive retrieval phase, user feedback is used to dynamically locate candidate images from clusters and outliers/outlier groups. Through relevance feedback, more information is gathered and fed to the learning algorithm-one-class SVM. Experiments show the effectiveness of the proposed platform

BACKGROUND: Protein sequence clustering has been widely used as a part of the analysis of protein structure and function. In most cases single linkage or graph-based clustering algorithms have been applied. OPTICS (Ordering Points To Identify the Clustering Structure) is an attractive approach due to its emphasis on visualization of results and support for interactive work, e.g., in choosing parameters. However, OPTICS has not been used, as far as we know, for protein sequence clustering. RESULTS: In this paper, a system of clustering proteins, SEQOPTICS (SEQuence clustering with OPTICS) is demonstrated. The system is implemented with Smith-Waterman as protein distance measurement and OPTICS at its core to perform protein sequence clustering. SEQOPTICS is tested with four data sets from different data sources. Visualization of the sequence clustering structure is demonstrated as well. CONCLUSION: The system was evaluated by comparison with other existing methods. Analysis of the results demonstrates that SEQOPTICS performs better based on some evaluation criteria including Jaccard coefficient, Precision, and Recall. It is a promising protein sequence clustering method with future possible improvement on parallel computing and other protein distance measurements.

MOTIVATION: In cluster analysis, the validity of specific solutions, algorithms, and procedures present significant challenges because there is no null hypothesis to test and no 'right answer'. It has been noted that a replicable classification is not necessarily a useful one, but a useful one that characterizes some aspect of the population must be replicable. By replicable we mean reproducible across multiple samplings from the same population. Methodologists have suggested that the validity of clustering methods should be based on classifications that yield reproducible findings beyond chance levels. We used this approach to determine the performance of commonly used clustering algorithms and the degree of replicability achieved using several microarray datasets. METHODS: We considered four commonly used iterative partitioning algorithms (Self Organizing Maps (SOM), K-means, Clutsering LARge Applications (CLARA), and Fuzzy C-means) and evaluated their performances on 37 microarray datasets, with sample sizes ranging from 12 to 172. We assessed reproducibility of the clustering algorithm by measuring the strength of relationship between clustering outputs of subsamples of 37 datasets. Cluster stability was quantified using Cramer's v2 from a kXk table. Cramer's v2 is equivalent to the squared canonical correlation coefficient between two sets of nominal variables. Potential scores range from 0 to 1, with 1 denoting perfect reproducibility. RESULTS: All four clustering routines show increased stability with larger sample sizes. K-means and SOM showed a gradual increase in stability with increasing sample size. CLARA and Fuzzy C-means, however, yielded low stability scores until sample sizes approached 30 and then gradually increased thereafter. Average stability never exceeded 0.55 for the four clustering routines, even at a sample size of 50. These findings suggest several plausible scenarios: (1) microarray datasets lack natural clustering structure thereby producing low stability scores on all four methods; (2) the algorithms studied do not produce reliable results and/or (3) sample sizes typically used in microarray research may be too small to support derivation of reliable clustering results. Further research should be directed towards evaluating stability performances of more clustering algorithms on more datasets specially having larger sample sizes with larger numbers of clusters considered.

Detecting outliers is an important topic in data mining. Sometimes the outliers are more interesting than the rest of the data. Outlier identification has lots of applications, such as intrusion detection, and unusual usage of credit cards or telecommunication services. In this paper, we propose a novel method for outlier identification which is based on network flow. We use the well known Maximum Flow Minimum Cut theorem from graph theory to find the outliers and strong outlier groups, and evaluate the outliers by the volume of the flow. This outlier detection occurs in a novel setting: to repair poor quality clusters generated by a clustering algorithm.

This paper introduces a system, called PolyCluster, which adopts state-of-the-art algorithms for data visualization and integrates human domain knowledge into the construction process of classification rules. By utilizing PolyCluster, users can obtain the visual representation for underlying datasets, and utilize that information to draw polygons to encompass wellformed clusters. Each polygon, along with its corresponding projection plane and associated attributes (or dimensions), will be saved as a classification rule, called a PolyRule, for later prediction tasks. Experimental evaluation shows that PolyCluster is a visual-based approach that offers numerous improvements over previous visual-based techniques. It also can help users to obtain additional knowledge from current datasets.

Frequent set generation is normally the first step toward association rule generation, and is usually considered the bottleneck step. Accordingly, many algorithms to generate frequent sets have been proposed. Evaluation of these algorithms is normally done empirically: their running times have been compared on a few or a substantial number of data files. Evaluation of the computational complexity of these algorithms has not much been pursued except to note that every algorithm requires exponential time, because the output size can be exponential in terms qf' the input size. We develop an alternate method qf' evaluating algorithms for frequent set generation: we propose evaluating the algorithms according to their computational complexity on several abstractly defined infinite families qf' data files. We perform this evaluation for Apriori and F P-growth on two infinite families qf'data files, and show that on both families, a variant of F P-growth is unboundedly faster than Apriori as the data file size increases.

OBJECTIVES: The authors consider the problem of identifying new, unexpected, and interesting patterns in hospital infection control and public health surveillance data and present a new data analysis process and system based on association rules to address this problem. DESIGN: The authors first illustrate the need for automated pattern discovery and data mining in hospital infection control and public health surveillance. Next, they define association rules, explain how those rules can be used in surveillance, and present a novel process and system--the Data Mining Surveillance System (DMSS)--that utilize association rules to identify new and interesting patterns in surveillance data. RESULTS: Experimental results were obtained using DMSS to analyze Pseudomonas aeruginosa infection control data collected over one year (1996) at University of Alabama at Birmingham Hospital. Experiments using one-, three-, and six-month time partitions yielded 34, 57, and 28 statistically significant events, respectively. Although not all statistically significant events are clinically significant, a subset of events generated in each analysis indicated potentially significant shifts in the occurrence of infection or antimicrobial resistance patterns of P. aeruginosa. CONCLUSION: The new process and system are efficient and effective in identifying new, unexpected, and interesting patterns in surveillance data. The clinical relevance and utility of this process await the results of prospective studies currently in progress.

This paper presents an $O( n\log n )$ algorithm for the bandwidth problem on interval graphs. Given an interval model for an interval graph with n vertices and an integer k, the algorithm constructs a layout of bandwidth at most k, if there exists one. Two previous algorithms for this problem have been published. One of them is flawed; the other is by Kleitman and Vohra and has complexity $O( nk )$.

The authors formalize the problem of minimizing the length of the longest interprocessor wire as the problem of embedding the processors of a hypercube onto a rectangular mesh, so as to minimize the length of longest wire. Where neighboring nodes of the mesh are taken as being at unit distance from one another, and where wires are constrained to be laid out as horizontal and vertical wires, the length of the wire joining nodes u and v of the mesh equals the graph-theoretic distance between u and v. The problem of minimizing delays due to interprocessor communication is then modeled as the problem of embedding the vertices of a hypercube onto the nodes of a mesh, so as to minimize dilation. Two embeddings which achieve dilations that (for large n) are within 26% of the lower bound for square meshes and within 12% for meshes with aspect ratio 2 are presented.>

Consideration is given to the performance of parallel best-bound-first branch-and-bound algorithms in which several nodes with least lower bounds are expanded simultaneously. It is well known that anomalies may occur in the execution of a parallel branch-and-bound algorithm. The authors show the conditions under which anomalies are guaranteed not to occur when the number of processors is doubled, or not even doubled.