We propose a novel framework based on Bayesian principles for validating clusterings and present efficient algorithms for use with centroid or exemplar based clustering solutions. Our framework treats the data as fixed and introduces perturbations into the clustering procedure. In our algorithms, we scale the distances between points by a random variable whose distribution is tuned against a baseline null dataset. The random variable is integrated out, yielding a soft assignment matrix that gives the behavior under perturbation of the points relative to each of the clusters. From this soft assignment matrix, we are able to visualize inter-cluster behavior, rank clusters, and give a scalar index of the the clustering stability. In a large test on synthetic data, our method matches or outperforms other leading methods at predicting the correct number of clusters. We also present a theoretical analysis of our approach, which suggests that it is useful for high dimensional data. / Science, Faculty of / Computer Science, Department of / Graduate

Researchers have been involved for decades in search of an efficient skin detection method. Yet current methods have not overcome the major limitations. To overcome these limitations, in this dissertation, a clustering and region growing based skin detection method is proposed. These methods together with a significant insight result in a more effective algorithm. The insight concerns a capability to define dynamically the number of clusters in a collection of pixels organized as an image. In clustering for most problem domains, the number of clusters is fixed a priori and does not perform effectively over a wide variety of data contents. Therefore, in this dissertation, a skin detection method has been proposed using the above findings and validated. This method assigns the number of clusters based on image properties and ultimately allows freedom from manual thresholding or other manual operations. The dynamic determination of clustering outcomes allows for greater automation of skin detection when dealing with uncertain real-world conditions.

Eastern Newts (Notophthalmus viridescens) are a ubiquitous member of eastern North America’s caudate fauna. Unlike the typical amphibian, their life cycle is split into three phases instead of two, commonly called a triphasic life cycle. The larvae of N. viridescens are fully aquatic, eventually metamorphosing to become terrestrial juveniles, called efts. Upon sexual maturity, the eft will metamorphose into a semi-aquatic adult where its external morphology is typical of an aquatic salamander. Since there are apparent differences in their ecological niche, there are different forces acting on their skeletons. We hypothesize that due to differences in buoyancy, torsion, and locomotion, differences are expected in the morphology of the axial skeleton. Using image data generated on a SkyScan 1273 micro-computed tomography (µCT) scanner, 3D shape analyses will be used to quantify shape differences between vertebrae and test the hypothesis. Three dimensional digital models of each vertebrae of interest will be rendered from the scans in Dragonfly (Object Research Systems). Each 3D model is then loaded into SlicerMorph (3D Slicer), where landmarks are placed upon homologous structures on each vertebra. A Generalized Procrustes Analysis (GPA) followed by a Principal Component Analysis (PCA) is conducted for each vertebra to test for potential shape differences between each life stage. GPA and PCA analysis will be conducted on 10 terrestrial juveniles, 10 semi-aquatic adults, 5 aquatic juveniles, and 5 paedomorphic adults. The 5 aquatic juveniles and 5 paedomorphic adults, eastern newts that remain in the water through their entire lives, will validate if the semi-aquatic adult is truly adapting towards an aquatic lifestyle. If GPA and PCA indicate statistical shape differences between certain vertebrae, those vertebrae will be run through the Automated Landmarking through Pointcloud Alignment and Correspondence Analysis (ALPACA) module of SlicerMorph to produce heatmap data on the 3D models showing where exactly the shape changes are occurring in the vertebra.

Wireless sensor networks (WSNs) are systems of resource-constrained sensor nodes (SNs), distributed throughout a sensor field. Energy limitations persist due to the wireless nature of SNs and an interest in minimizing the cost and physical footprint of SNs. Due to the resource-constrained nature of SNs, much WSN research has focused on energy-efficient communication algorithms. Communication algorithms are necessary for energy-efficient data transmission between SNs and the transmission of data collected by SNs to a base station. A popular algorithm known as Low-Energy Adaptive Clustering Hierarchy (LEACH) achieves more energy-efficient communication by organizing SNs into clusters for localized communication. When SNs are mobile, the energy efficiency of LEACH is degraded because of geographic dispersion of SN clusters. This thesis proposes LEACH-Centered Cluster-head (LEACH-CCH), a clustering algorithm aimed at improving WSN lifetime in cases of stationary and mobile sensor nodes. Mobile sensor network applications are explored including vehicle-to-infrastructure communication networks.

Pyroprinting is a novel, library-based microbial source tracking method developed by the Biology department at Cal Poly, San Luis Obispo. This method consists of two parts: (1) a collection of bacterial fingerprints, called pyroprints, from known host species, and (2) a method for pyroprint comparison. Currently, Cal Poly Library of Pyroprints (CPLOP), a web-based database application, provides storage and analysis of over $10000$ pyroprints. This number is quickly growing as students and researchers continue to use pyroprinting for research. Biologists conducting research using pyroprinting rely on methods for partitioning collected bacterial isolates into bacterial strains. Clustering algorithms are commonly used for bacterial strain analysis of organisms in computational biology. Unfortunately, agglomerative hierarchical clustering, a commonly used clustering algorithm, is inadequate given the nature of data collection for pyroprinting. While the clusters produced by agglomerative hierarchical clustering are acceptable, pyroprinting requires a method of analysis that is scalable and incorporates useful metadata into the clustering process. We propose ontology-based hierarchical clustering (OHClust!), a modification of agglomerative hierarchical clustering that expresses metadata-based relationships as an ontology to direct the order in which hierarchical clustering algorithms analyze the data. In this thesis, the strengths and weaknesses of OHClust! are discussed, and its performance is analyzed in comparison to agglomerative hierarchical clustering.

Since supplying passenger safety is the fundamental purpose of establishing vehicular ad hoc networks, researchers are mainly concerned with giving a strategy to ensure meeting this demand. Rescue services vehicles are one of the primary components of VANET, and the messages they broadcast play a vital part in delivering safety. As a result, offering a routing protocol algorithm that prioritizes their messages and guarantees QoS is one approach to meet the core purpose of VANET, or in other words, safety. Grouping the vehicle nodes, also known as clustering, is one of the appropriate solutions for improving the performance of the routing protocol. Most traditional cluster-based topologies in VANET solely consider the mobility to pick the cluster head. While cluster head speed fluctuations can affect the number of cluster head choices and improve the VANET QoS. Furthermore, the fast movement of vehicles exposes clustering links to vulnerability and directly impacts reliability. In this research, we proposed a new efficient multi-hop cluster-based routing protocol that prioritizes emergency vehicles' messages and provides stable and reliable communications to ensure emergency messages are delivered effectively. The number of neighbors, the distance between the vehicle and BS, and the S/N power received by the base station (BS) are the criteria for selecting a cluster head (CH) in our proposed model. Each CH is in charge of transmitting data from other cluster members to the infrastructure directly or indirectly through the qualified adjacent cluster heads. The epigenomic workflow approach is employed in our design to establish efficient resource management in RSU computing nodes and decrease transmission latency. The transmitted messages from vehicles are divided into two separate queues (q1 and q2), while the priority message queue uses the TDMA protocol to disseminate. In this thesis, experiments and simulations were carried out utilizing the NS-2 tool with VanetMobisim integration to assess the efficacy of the suggested approach. Achieved results indicate that the proposed technique increases communication reliability and stability while improving QoS for messages provided by rescue services.

No description available.

No description available.

No description available.

Clustering procedures suitable for the analysis of very high-dimensional data are needed for many modern data sets. One model-based clustering approach called high-dimensional data clustering (HDDC) uses a family of Gaussian mixture models to model the sub-populations of the observed data, i.e., to perform cluster analysis. The HDDC approach is based on the idea that high-dimensional data usually exists in lower-dimensional subspaces; as such, the dimension of each subspace, called the intrinsic dimension, can be estimated for each sub-population of the observed data. As a result, each of these Gaussian mixture models can be fitted using only a fraction of the total number of model parameters. This family of models has gained attention due to its superior classification performance compared to other families of mixture models; however, it still suffers from the usual limitations of Gaussian mixture model-based approaches. Herein, a robust analogue of the HDDC approach is proposed. This approach, which extends the HDDC procedure to include the mulitvariate-t distribution, encompasses 28 models that rectify one of the major shortcomings of the HDDC procedure. Our tHDDC procedure is fitted to both simulated and real data sets and is compared to the HDDC procedure using an image reconstruction problem that arose from satellite imagery of Mars' surface. / Thesis / Master of Science (MSc)