Recent Advances in Modularity Optimization and Their Application in Retailing

The MR brain image analysis is used to extract clinical information that improve diagnosis and treatment of disease. Brain tumors are one of the most common brain diseases. Clustering is a process for segmenting and classifying objects. There are many clustering strategies such as the hard clustering scheme and the fuzzy clustering scheme, each of them has its own special characteristics. In this work Fuzzy C-Mean clustering was implemented to segment three abnormal brain MR images, and the performance of it was analyised. This algorithms was applied to cluster the images into different clusters number: 5-9 with different values of membership grade: 0.50-0.90 with steps of 0,05 for each cluster number. The percentage of the unclassified pixels that were produced from implementing FCM algorithm with different configuration was calculated. The minimum values of the objective function of the FCM algorithm for different number of clusters and for different membership grade values were also calculated. The results showed that an optimal number of clusters that corresponds to optimal segmentation error is depending on the slice condition. In this experiment, the optimal cluster number was found to be 6. The fluctuation around this number is affected also by the anatomical structure of the slice. In addition, it can be concluded that the objective function may not be the superior criterion for the judgments of goodness, where it may be a few number of pixels with high uncertainty is the source of high error.

Cluster computing has recently gone through an evolution from single processor systems to multicore/multi-socket systems. This has resulted in lowering the cost/performance ratio of the compute machines. Compute farms that host these machines tend to become heterogeneous over time due to incremental extensions, hardware upgrades and/or nodes being purchased for users with particular needs. This heterogeneity is not surprising given the wide range of processor, memory and network technologies that become available and the relatively small price difference between these various options. Different CPU architectures, memory capacities, communication and I/O interfaces of the participating compute nodes present many challenges to job scheduling and often result in under or over utilization of the compute resources. In general, it is not feasible for the application programmers to specifically optimize their programs for such a set of differing compute nodes, due to the difficulty and time-intensiveness of such a task. The trend of heterogeneous compute farms has coincided with resurgence in the virtualization technology. Virtualization technology is receiving widespread adoption, mainly due to the benefits of server consolidation and isolation, load balancing, security and fault tolerance. Virtualization has also generated considerable interest in the High Performance Computing (HPC) community, due to the resulting high availability, fault tolerance, cluster partitioning and accommodation of conflicting user requirements. However, the HPC community is still wary of the potential overheads associated with virtualization, as it results in slower network communications and disk I/O, which need to be addressed. The live migration feature, available to most virtualization technologies, can be leveraged to improve the throughput of a heterogeneous compute farm (HC) used for HPC applications. For this we mitigated the slow network communication in Xen; an open source virtual machine monitor. We present a detailed analysis of the communication framework of Xen and propose communication configurations that give 50% improvement over the conventional Xen network configuration. From a detailed study of the migration facility in Xen, we propose an improvement in the live migration facility specifically targeting HPC applications. This optimization gives around 50% improvement over the default migration facility of Xen. In this thesis, we also investigate resource scheduling in heterogeneous compute farm with the perspective of dynamic resource re-mapping. Our approach is to profile each job in the compute farm at runtime, and propose a better resource mapping compared to the initial allocation. We then migrate the job(s) to the best-suited homogeneous sub-cluster to improve overall throughput of the HC. For this, we develop a novel heterogeneity and virtualization-aware profiling framework, which is able to predict the CPU and communication characteristics of high performance scientific applications. Our experimental setup consists of a 32 node virtualized cluster with machines having CPU frequency between 2 GHz to 3 GHz. The framework was tested with a subset of NAS parallel benchmarks with the workload generated from Dror Feitelson’s workload archives. The prediction accuracy of our performance estimation model is over 80%. The framework implementation is lightweight, with an overhead of 3%. Our experiments show that we are able to improve the throughput of the compute farm by 25% and the time saved by the HC with our framework is over 30%. The framework can be readily extended to HCs supporting a cloud computing environment.

The study was aimed at assessing the genetic variation and relatedness among four Capsicum varieties through electrophoretic separation of their leaf and seed proteins by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE). Total seed and leaf proteins were extracted and separated on 12% polyacrylamide gels using standard protocols. Distinct polymorphism in electrophoretic banding patterns of seed and leaf proteins was observed in the four Capsicum varieties through a total of thirty-eight (38) polypeptide bands in the seeds and a total of seventeen (17) polypeptide bands in the leaves analyzed. Variation existed not only in the number of bands but also in the intensity of bands in both the leaf and seed samples studied. Sokal and Sneath's coefficient of similarity revealed a generally high level of similarity in the seed protein bands of the four varieties studied ranging from 50% to 100%. It also revealed an average level of similarity in the leaf protein bands of the four varieties of Capsicum studied ranging from 16.7% to 83.3%. A dendrogram constructed based on the Single Linkage Cluster Analysis(SCLA) using the relative mobility values of seed and leaf proteins of the four varieties revealed two major clusters. From the seed, cluster 1 contains (C. annuum var. abbreviatum and C. annuum var. acuminatum), and cluster 2(C. annuum var. grossum, and C.frutescens var.baccatum). Also, two prominent clusters were revealed in the leaf sample, with cluster 1 containing (C. annuum var. abbreviatum, C. annuum var. acuminatum and C.frutescens var.baccatum) and cluster 2(C. annuum var. grossum). Significance of the findings has been discussed. Plant breeders can use the information on variation and relatedness among the Capsicum species for cultivar development in pepper.

A Mobile Ad-Hoc Network (MANET) is a collection of wireless mobile nodes that is self-organized to form an arbitrary topology with the absence of fixed infrastructure or centralized administration .In such environment security has become a main concern to provide confidentiality, authentication, integrity, availability, access control, and non-repudiation. Key management is the main component in MANET security.This paper proposes an efficient and secure distributed group key management scheme. Where, nodes in the network are divided into clusters. Each cluster has a cluster head (CH) and cluster members. Cluster head is responsible for maintaining the group key. It also updates the group key whenever a member joins or leaves the cluster. There-keying process takes place within the cluster only. So the computation and communication cost will be reduced. Another level of security is

E 01 Modeling Stellar Collisions in Star Clusters E 02 Fast Modelling of Individual Star Clusters E 03 Star Clusters Surviving Low Star Formation Efficiencies E 04 Extreme Stellar Dynamics: Collisions and Captures in Dense Stellar Clusters E 05 Dense Stellar Systems as Tracers of the Formation and Evolution of Galaxies: Advanced Tools for the Determination of Age and Metallicity of Old Globular Clusters E 06 The Merging and Tidal Destruction of Globular Clusters within Inner Galactic Regions E 07 Galaxy Models with Central Density Cusp E 08 The Formation of Super-Star Clusters E 09 Gravitational Radiation from Black Hole Triple Systems E 10 Structure in the Mass Function of Star Clusters E 11 The 30 Doradus Starburst Cluster: Infrared Luminosity Function and Low-mass IMF in a Spatially Resolved Dense Young Stellar System E 12 Brown Dwarfs from Decaying Accreting Triple Systems E 13 Birth and Evolution of Stars in AGN Disks E 14 Formation and Evolution of Clumpy Tidal Tails Around Globular Clusters E 15 Blue Stragglers in Galactic Globular Clusters: Tracing the Effect of Dynamics on Stellar Evolution E 16 Massive, Dense Star Clusters in the Nuclei of Bulge-less Spiral Galaxies

To deal with the limitation that flat routing can hardly be accustomed to large scale Underwater Sensor Networks (USN), a new clustering routing algorithm Dynamic Layered Clustering Routing (DLCR) is proposed, which can be accustomed to larger scale networks. This algorithm divides the networks into several layers from top to bottom, and selects the nodes which have more remaining energy and shorter distance to sink as the cluster head nodes, thus, clusters communication energy consumption are reduced. In order to avoid the same nodes being elected to be cluster head nodes continuously, a dynamic layered mechanism that the networks are divided into different layers in each circle of data gathering is proposed. The experiment shows that DLCR not only has a better stability, but also reduces the energy consumption and prolongs the lifetime of the whole networks.

Wireless sensor networks (WSNs) are used for various applications, such as habitat monitoring, automation, agriculture, and security. In order to send information from very high number of sensor nodes to the base station, it is necessary and economical to group sensors into clusters to prolong network lifetime. Due to the resource limitation of sensor nodes, the collected information from sensor nodes in the cluster has to be compressed quickly and precisely for transmission. In this paper, we propose a VQ-LBG based approach for cluster formation in WSN. The most distinguishing feature of the proposed method is that both energy efficient cluster formation and fast data compression can be guaranteed. Experiment shows its great improvement over other related methods.

A novel feature extraction method that utilizes nonlinear mapping from the original data space to the feature space is presented in this dissertation. Feature extraction methods aim to find compact representations of data that are easy to classify. Measurements with similar values are grouped to same category, while those with differing values are deemed to be of separate categories. For most practical systems, the meaningful features of a pattern class lie in a low dimensional nonlinear constraint region (manifold) within the high dimensional data space. A learning algorithm to model this nonlinear region and to project patterns to this feature space is developed. Least squares estimation approach that utilizes interdependency between points in training patterns is used to form the nonlinear region. The proposed feature extraction strategy is employed to improve face recognition accuracy under varying illumination conditions and facial expressions. Though the face features show variations under these conditions, the features of one individual tend to cluster together and can be considered as a neighborhood. Low dimensional representations of face patterns in the feature space may lie in a nonlinear constraint region, which when modeled leads to efficient pattern classification. A feature space encompassing multiple pattern classes can be trained by modeling a separate constraint region for each pattern class and obtaining a mean constraint region by averaging all the individual regions. Unlike most other nonlinear techniques, the proposed method provides an easy intuitive way to place new points onto a nonlinear region in the feature space. The proposed feature extraction and classification method results in improved accuracy when compared to the classical linear representations. Face recognition accuracy is further improved by introducing the concepts of modularity, discriminant analysis and phase congruency into the proposed method. In the modular approach, feature components are extracted from different sub-modules of the images and concatenated to make a single vector to represent a face region. By doing this we are able to extract features that are more representative of the local features of the face. When projected onto an arbitrary line, samples from well formed clusters could produce a confused mixture of samples from all the classes leading to poor recognition. Discriminant analysis aims to find an optimal line orientation for which the data classes are well separated. Experiments performed on various databases to evaluate the performance of the proposed face recognition technique have shown improvement in recognition accuracy, especially under varying illumination conditions and facial expressions. This shows that the integration of multiple subspaces, each representing a part of a higher order nonlinear function, could represent a pattern with variability. Research work is progressing to investigate the effectiveness of subspace projection methodology for building manifolds with other nonlinear functions and to identify the optimum nonlinear function from an object classification perspective.

The minimum rate requirements of users are not taken into consideration in the existing power allocation schemes for maximum fairness of non-orthogonal multiple access (NOMA) systems. For this problem, the poa power allocation scheme for fairly improving user rate in NOMA systems is proposed. Firstly, the minimum power required for each cluster is calculated based on channel conditions and rate requirement for each user. Secondly, in the case of meeting the minimum rate requirements of all users, the power allocation optimization problem of improving each user’s rate fairly is established with the constraints of minimum power required by each cluster and the total power required by all clusters. Finally, the power allocation scheme, which satisfies the minimum rate demand of the user and maximizes the user's improved minimum rate, is obtained by adjusting the power of the part clusters. The simulation results show that when the minimum rate requirements of users are different, the increased rates of users and the outage probabilities of the proposed scheme both outperform these of the existing schemes in the same scenario.

For the description of scattering processes and resonances a proper treatment of the continuum is necessary. To achieve such a description in a microscopic many-body approach one has to connect compact configurations that describe the internal part of the wave function with external cluster configurations representing the open channels. We have developed such an approach within fermionic molecular dynamics (FMD). FMD uses a wave-packet basis that allows to describe the internal parts of the wave function and the external cluster channels on equal footing. The matching to the asymptotic behavior as given by two point-like clusters interacting only via Coulomb is performed in the microscopic R-matrix formalism. As a first example we discuss the continuum states in C. We previously studied C in bound state approximation [1]. The focus was on the properties of the Hoyle state, which lies just above the Be-He threshold. Whereas the bound state approximation is expected to work well for a very narrow resonance like the Hoyle state this is no longer true for other resonances like the second 2 state. The resonance position and width for this state could only be determined very recently by direct excitation with photons [2]. The existence and the nature of other states in the continuum is still hotly debated. To address these questions we extend our calculations with a proper treatment of the continuum. Before doing the full FMD calculation a study within the microscopic αcluster model has been performed. The microscopic cluster model with full antisymmetrization and employing a phenomenological two-body interaction proved to be successful in describing many properties of C. In the internal region the Hilbert space is built from three-α configurations on a triangular grid. In the external region Be-He configurations are added. The Be eigenstates are obtained in bound state approximation by diagonalizating configurations up to 9 fm distance. Our results will have to be checked for convergence with respect to increasing the number of included Be (pseudo-) states. Technically the challenge is related to the restoration of rotational symmetry. First the intrinsic wave functions for Be have to be projected on good angular momentum. In a second step the Be-He configurations (with different orientations of the Be spin) have to be projected on total angular momentum. For solving the Schrodinger equation with the microscopic R-matrix method, Hamiltonian and norm kernels for the different channels have to be calculated. In Fig. 1 the calculated phase shifts (from the diagonal

Increasing customer bandwidth needs to be considered by companies by determining potential customers. Potential determination is needed because previously it was only done randomly. Therefore, first determining the potential is needed by grouping customers who have the same characteristics based on the data and attributes they have. This research will implement data mining technique with clustering method with K-means algorithm on 263 FTTH Broadband customer groups. Then the potency can be determined based on the final centroid point in the grouping. The results obtained are divided into 5 clusters consisting of 34 customers or 12.92% of the total potential customers, 29 customers or 11.02% of the total potential customers, 56 customers or 21.30% of the total potential customers, 54 customers. or 20.53% of the total fewer potential customers and 90 customers or 34.22% of the total non-potential customers. Comparison of the validity of the Davies-Bouldin Index cluster for 5 groups between K-Means and K-Medoids, the value of 0.538 for K-Means and 0.819 for K-Medoids is obtained. This method is used to streamline the distribution of bandwidth.

This study seeks to capture the mango value preferences of consumers in Pakistan with a view to identifying potential improvements in value chain practices in the mango industry. Data were collected through five focus group discussions and an intercept survey of 450 consumers in four cities of Pakistan. Three consumer clusters were identified through cluster analysis. These clusters had significant differences in terms of their search, experience, safety and marketing attributes. Given the existing consumer value preferences and likely changes in consumer behavior, the study suggested that value chain actors can improve their practices so as to improve consumer value and achieve higher returns.

This work was carried out on travelers Pereira city until 2014. With the collected data segmentation and description of travel behavior performing cluster analysis was performed subsequently classified according to trends and characteristics applying association rules and eventually were grouped by the level of expenditure, perception and satisfaction after visiting the city through cluster analysis. In the results they yielded 5 clusters from which one it was realized the analysis.

This article extends previous empirical research on the benefits and barriers for researchers in the field of Information Systems (IS) to engage in University-Industry Collaboration (UIC), a term that is used to describe the active engagement of a company or government agency in a joint research project with academics. The objectives and motivations of UIC have been discussed widely and controversially in the literature. We were particularly interested in the perceived benefits and barriers that influence an individual researcher to engage (or not engage) in a joint research project. An in-depth literature analysis showed that a very small number of research articles contain empirical data and that there are serious methodological issues. In order to address these issues, we collected primary data from 328 IS researchers with experience in UIC and conducted an explorative study. We developed parsimonious formative measures for the benefits and barriers of UIC and we found that academic and economic benefits positively influence the intention to conduct UIC in the future, while economic barriers negatively influence the intention to engage in UIC. A cluster analysis found five clusters (groups of researchers) that differ in their perception of benefits and barriers and the future intention to conduct UIC. However, the majority of the researchers have a very high intention to conduct UIC in the future.

Early events in folding can determine if a protein is going to fold, misfold, or aggregate. Understanding these deterministic events is paramount for de novo protein engineering, the enhancement of biopharmaceutical stabilities, and understanding neurodegenerative diseases including amyotrophic lateral sclerosis and Alzheimer's disease. However, the physicochemical and structural biases within high energy states of protein biopolymers are poorly understood. A combined experimental and computational study was conducted on the small β/α-repeat protein CheY to determine the structural basis of its submillisecond misfolding reaction to an off-pathway intermediate. Using permutations, we were able to discriminate between the roles of two proposed mechanisms of folding; a nucleation condensation model, and a hydrophobic collapse model driven by the formation of clusters of isoleucine, leucine, and valine (ILV) residues. We found that by altering the ILV cluster connectivity we could bias the early folding events to either favor on or off-pathway intermediates. Structural biases were also experimentally observed in the unfolded state of a de novo designed synthetic β/α-repeat protein, Di-III_14. Although thermodynamically and kinetically 2-state, Di-III_14 has a well structured unfolded state that is only observable under native-favoring conditions. This unfolded state appears to retain native-like structure, consisting of a hydrophobic 7 core (69% ILV) stabilized by solvent exposed polar groups and long range electrostatic interactions. Together, these results suggest that early folding events are largely deterministic in these two systems. Generally, low contact order ILV clusters favor local compaction and, in specific cases, long range electrostatic interactions may have stabilizing effects in higher energy states.