The task of multi-body motion segmentation refers to segmenting feature trajectories in a sequence of images according to their 3D motion affinity without knowing the number of motions in advance. It is critical for understanding and reconstructing a dynamic scene. This problem essentially consists of two subproblems, segmenting features and detecting the number of motions. While the state-of-the-art LBF algorithm achieves segmentation accuracy as high as 96.5%, it is still disturbed by a phenomenon called over-locality. A novel mode seeking algorithm with an adaptive distance measure is proposed to avoid this problem, and improves the accuracy to 98.1%. The LBF algorithm is incapable of detecting the number of motions itself. A randomized version of the mode seeking algorithm is presented, which could detect the number as well as preserve satisfactory segmentation accuracy. To detect the number of motions, a kernel optimization method locates it via kernel alignment. However, it suffers from over-locality and over-detects the number of motions. An intersection measure and two mutual information measures are presented to solve this problem. Using these measures, the proposed clustering consensus framework recasts the motion number detection problem to a clustering consensus problem. It extends the kernel optimization method from two-clustering consensus to multiple-clustering consensus. A large number of experiments and comparisons have been done, and convincing results are obtained. / published_or_final_version / Computer Science / Doctoral / Doctor of Philosophy

Not available / text

Not available / text

published_or_final_version / Physics / Master / Master of Philosophy

Single nucleotide polymorphisms (SNPs) have been increasingly popular for a wide range of genetic studies. A high-throughput genotyping technologies usually involves a statistical genotype calling algorithm. Most calling algorithms in the literature, using methods such as k-means and mixturemodels, rely on elliptical structures of the genotyping data; they may fail when the minor allele homozygous cluster is small or absent, or when the data have extreme tails or linear patterns. We propose an automatic genotype calling algorithm by further developing a linear grouping algorithm (Van Aelst et al., 2006). The proposed algorithm clusters unnormalized data points around lines as against around centroids. In addition, we associate a quality value, silhouette width, with each DNA sample and a whole plate as well. This algorithm shows promise for genotyping data generated from TaqMan technology (Applied Biosystems). A key feature of the proposed algorithm is that it applies to unnormalized fluorescent signals when the TaqMan SNP assay is used. The algorithm could also be potentially adapted to other fluorescence-based SNP genotyping technologies such as Invader Assay. Motivated by the SNP genotyping problem, we propose a partial likelihood approach to linear clustering which explores potential linear clusters in a data set. Instead of fully modelling the data, we assume only the signed orthogonal distance from each data point to a hyperplane is normally distributed. Its relationships with several existing clustering methods are discussed. Some existing methods to determine the number of components in a data set are adapted to this linear clustering setting. Several simulated and real data sets are analyzed for comparison and illustration purpose. We also investigate some asymptotic properties of the partial likelihood approach. A Bayesian version of this methodology is helpful if some clusters are sparse but there is strong prior information about their approximate locations or properties. We propose a Bayesian hierarchical approach which is particularly appropriate for identifying sparse linear clusters. We show that the sparse cluster in SNP genotyping datasets can be successfully identified after a careful specification of the prior distributions.

Food webs are good ecological macro-descriptors and their study is important in ecology in understanding nutrient cycles, tracing and quantifying energy and in describing trophic interactions within an ecosystem. The knowledge of food web finds applications in various natural sciences disciplines but also in many productive sectors. This study investigated the structure of the food web of a shallow lake in the Pantanal flood plain. The food web included two macrophytes, six aquatic insects, four crustaceans and 24 fish species. Sources of carbon for the various organisms living in the lake were identified through the values of δ13C exhibited by the organisms. The δ15N signature was used to estimate the trophic position of each organism. A cluster analysis based on the two isotopic signatures revealed six different feeding guilds and emphasized on the broad occurrence of omnivory among animals living in the lake. This study revealed that the use of food carbon was the most important factor that structured the lake community. Very low values of δ13C in zooplankton, benthic dwellers and bottom-feeder organisms as well as similarities between the gradient of δ13C and that of use of methane oxidizing bacteria informed on the possible use of biogenic methane as a source carbon and energy for the lake biota.

No description available.

Pervasive Developmental Disorders (PDDs) are neurodevelopmental disorders characterized by impairments in social interaction, communication and behaviour [Str04]. Given the diversity and varying severity of PDDs, diagnostic tools attempt to identify homogeneous subtypes within PDDs. The diagnostic system Diagnostic and Statistical Manual of Mental Disorders - Fourth Edition (DSM-IV) divides PDDs into five subtypes. Several limitations have been identified with the categorical diagnostic criteria of the DSM-IV. The goal of this study is to identify putative subtypes in the multidimensional data collected from a group of patients with PDDs, by using cluster analysis. Cluster analysis is an unsupervised machine learning method. It offers a way to partition a dataset into subsets that share common patterns. We apply cluster analysis to data collected from 358 children with PDDs, and validate the resulting clusters. Notably, there are many cluster analysis algorithms to choose from, each making certain assumptions about the data and about how clusters should be formed. A way to arrive at a meaningful solution is to use consensus clustering to integrate results from several clustering attempts that form a cluster ensemble into a unified consensus answer, and can provide robust and accurate results [TJPA05]. In this study, using cluster analysis, cluster validation, and consensus clustering, we identify four clusters that are similar to – and further refine  three of the five subtypes defined in the DSM-IV. This study thus confirms the existence of these three subtypes among patients with PDDs. / Thesis (Master, Computing) -- Queen's University, 2007-11-15 23:34:36.62 / OGS, QGA

Single nucleotide polymorphisms (SNPs) have been increasingly popular for a wide range of genetic studies. A high-throughput genotyping technologies usually involves a statistical genotype calling algorithm. Most calling algorithms in the literature, using methods such as k-means and mixturemodels, rely on elliptical structures of the genotyping data; they may fail when the minor allele homozygous cluster is small or absent, or when the data have extreme tails or linear patterns. We propose an automatic genotype calling algorithm by further developing a linear grouping algorithm (Van Aelst et al., 2006). The proposed algorithm clusters unnormalized data points around lines as against around centroids. In addition, we associate a quality value, silhouette width, with each DNA sample and a whole plate as well. This algorithm shows promise for genotyping data generated from TaqMan technology (Applied Biosystems). A key feature of the proposed algorithm is that it applies to unnormalized fluorescent signals when the TaqMan SNP assay is used. The algorithm could also be potentially adapted to other fluorescence-based SNP genotyping technologies such as Invader Assay. Motivated by the SNP genotyping problem, we propose a partial likelihood approach to linear clustering which explores potential linear clusters in a data set. Instead of fully modelling the data, we assume only the signed orthogonal distance from each data point to a hyperplane is normally distributed. Its relationships with several existing clustering methods are discussed. Some existing methods to determine the number of components in a data set are adapted to this linear clustering setting. Several simulated and real data sets are analyzed for comparison and illustration purpose. We also investigate some asymptotic properties of the partial likelihood approach. A Bayesian version of this methodology is helpful if some clusters are sparse but there is strong prior information about their approximate locations or properties. We propose a Bayesian hierarchical approach which is particularly appropriate for identifying sparse linear clusters. We show that the sparse cluster in SNP genotyping datasets can be successfully identified after a careful specification of the prior distributions.

Clustering is the search for those partitions that reflect the structure of an object set. Traditional clustering algorithms search only a small sub-set of all possible clusterings (the solution space) and consequently, there is no guarantee that the solution found will be optimal. We report here on the application of Genetic Algorithms (GAs) -- stochastic search algorithms touted as effective search methods for large and complex spaces -- to the problem of clustering. GAs which have been made applicable to the problem of clustering (by adapting the representation, fitness function, and developing suitable evolutionary operators) are known as Genetic Clustering Algorithms (GCAs). There are two parts to our investigation of GCAs: first we look at clustering into a given number of clusters. The performance of GCAs on three generated data sets, analysed using 4320 differing combinations of adaptions, establishes their efficacy. Choice of adaptions and parameter settings is data set dependent, but comparison between results using generated and real data sets indicate that performance is consistent for similar data sets with the same number of objects, clusters, attributes, and a similar distribution of objects. Generally, group-number representations are better suited to the clustering problem, as are dynamic scaling, elite selection and high mutation rates. Independent generalised models fitted to the correctness and timing results for each of the generated data sets produced accurate predictions of the performance of GCAs on similar real data sets. While GCAs can be successfully adapted to clustering, and the method produces results as accurate and correct as traditional methods, our findings indicate that, given a criterion based on simple distance metrics, GCAs provide no advantages over traditional methods. Second, we investigate the potential of genetic algorithms for the more general clustering problem, where the number of clusters is unknown. We show that only simple modifications to the adapted GCAs are needed. We have developed a merging operator, which with elite selection, is employed to evolve an initial population with a large number of clusters toward better clusterings. With regards to accuracy and correctness, these GCAs are more successful than optimisation methods such as simulated annealing. However, such GCAs can become trapped in local minima in the same manner as traditional hierarchical methods. Such trapping is characterised by the situation where good (k-1)-clusterings do not result from our merge operator acting on good k-clusterings. A marked improvement in the algorithm is observed with the addition of a local heuristic.