Face recognition is the problem of identifying individuals in images. This thesis evaluates two methods used to determine if pairs of face images belong to the same individual or not. The first method is a combination of principal component analysis and a neural network and the second method is based on state-of-the-art convolutional neural networks. They are trained and evaluated using two different data sets. The first set contains many images with large variations in, for example, illumination and facial expression. The second consists of fewer images with small variations. Principal component analysis allowed the use of smaller networks. The largest network has 1.7 million parameters compared to the 7 million used in the convolutional network. The use of smaller networks lowered the training time and evaluation time significantly. Principal component analysis proved to be well suited for the data set with small variations outperforming the convolutional network which need larger data sets to avoid overfitting. The reduction in data dimensionality, however, led to difficulties classifying the data set with large variations. The generous amount of images in this set allowed the convolutional method to reach higher accuracies than the principal component method.

High-speed transmission over contemporary communication networks has drawn many research efforts. Traffic scheduling schemes which play a critical role in managing network transmission have been pervasively studied and widely implemented in various practical communication networks. In a sophisticated communication system, a variety of applications co-exist and require differentiated Quality-of-Service (QoS). Innovative scheduling schemes and hybrid scheduling disciplines which integrate multiple traditional scheduling mechanisms have emerged for QoS differentiation. This study aims to develop novel analytical models for commonly interested scheduling schemes in communication systems under more realistic network traffic and use the models to investigate the issues of design and development of traffic scheduling schemes. In the open literature, it is commonly recognized that network traffic exhibits self-similar nature, which has serious impact on the performance of communication networks and protocols. To have a deep study of self-similar traffic, the real-world traffic datasets are measured and evaluated in this study. The results reveal that selfsimilar traffic is a ubiquitous phenomenon in high-speed communication networks and highlight the importance of the developed analytical models under self-similar traffic. The original analytical models are then developed for the centralized scheduling schemes including the Deficit Round Robin, the hybrid PQGPS which integrates the traditional Priority Queueing (PQ) and Generalized Processor Sharing (GPS) schemes, and the Automatic Repeat reQuest (ARQ) forward error control discipline in the presence of self-similar traffic. Most recently, research on the innovative Cognitive Radio (CR) techniques in wireless networks is popular. However, most of the existing analytical models still employ the traditional Poisson traffic to examine the performance of CR involved systems. In addition, few studies have been reported for estimating the residual service left by primary users. Instead, extensive existing studies use an ON/OFF source to model the residual service regardless of the primary traffic. In this thesis, a PQ theory is adopted to investigate and model the possible service left by selfsimilar primary traffic and derive the queue length distribution of individual secondary users under the distributed spectrum random access protocol.

published_or_final_version / Computer Science / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

Pore-scale network modeling has become an effective method for accurate prediction and upscaling of macroscopic properties, such as permeability. Networks are either mapped directly from real media or stochastic methods are used that simulate their heterogeneous pore structure. Flow is then modeled by enforcing conservation of mass in each pore and approximations to the momentum equations are solved in the connecting throats. In many cases network modeling compares favorably to experimental measurements of permeability. However, computational and imaging restrictions generally limit the network size to the order of 1 mm3 (few thousand pores). For extremely heterogeneous media these models are not large enough in capturing the petrophysical properties of the entire heterogeneous media and inaccurate results can be obtained when upscaling to the continuum scale. Moreover, the boundary conditions imposed are artificial; a pressure gradient is imposed in one dimension so the influence of flow behavior in the surrounding media is not included. In this work we upscale permeability in large, heterogeneous media using physically-representative pore-scale network models (domain ~106 pores). High-performance computing is used to obtain accurate results in these models, but a more efficient, novel domain decomposition method is introduced for upscaling the permeability of pore-scale models. The medium is decomposed into hundreds of smaller networks (sub-domains) and then coupled with the surrounding models to determine accurate boundary conditions. Finite element mortars are used as a mathematical tool to ensure interfacial pressures and fluxes are matched at the interfaces of the networks boundaries. The results compare favorably to the more computationally intensive (and impractical) approach of upscaling the media as a single model. Moreover, the results are much more accurate than traditional hierarchal upscaling methods. This upscaling technique has important implications for using pore-scale models directly in reservoir simulators in a multiscale setting. The upscaling techniques introduced here on single phase flow can also be easily extended to other flow phenomena, such as multiphase and non-Newtonian behavior. / text

People always find themselves interacting with others and forming ties with them; these ties shape an individual’s social network which helps form the self-conception and identity of a person. In discussing the essence of social networks and how they are formed the concept of homophily is of high significance. Therefore, the purpose of this research is to show the association between homophily and the process of friendship formation. As the structure of any social network is important in tie formation, I have also intended to study homophilous tie formation from a distinctiveness theory perspective, suggesting that individuals with minority characteristics are more prone to form friendship ties with each other. The types of homophily studied in this research are gender, religion, nationality/ethnicity, and political views. The data is gathered from the cohort which started grade 10 in upper secondary education in a school in Stockholm in Autumn 2012. The analyses were conducted using logistic regression. The results indicated the existence of gender homophily and national homophily. However, religious homophily did not appear to be significant; political homophily was only significant for individuals who were participating in political meetings. However, due to lack of sufficient data, the relations between network structure and homophilous relations could not be accurately tested. Since the data were not collected randomly and the school was chosen due to its specific characteristics, it is not possible to generalize the results of the research to all of the adolescents living in Stockholm. However, this research sheds some light on the mechanisms at play in friendship formation among adolescents.

Following advancements in the "omics" fields of molecular biology and genetics, much attention has been focused on categorising and annotating the large volume of data that has been produced since the sequencing of human and model genomes. With high-throughput data generated from these "omics" experiments and the increasing deposition of information from genetics experiments in biological databases, our understanding of the mechanisms that bridge the gap from genotype to phenotype can be explored in a holistic context. This is one of the aims of the relatively new field of systems biology, which aims to understand the complexity of biological systems in a holistic manner by studying the system as an ensemble of interacting parts. With increased volume and comprehensiveness of biological data, prediction of gene function and automatic identification of potential models for human diseases have become important aspects of systems-level analysis for wet-lab geneticists and clinicians. Here, I describe an integrated analysis of mouse phenotype data with high-throughput experiments to give genome-wide information about gene relationships and their function in a systems biology context. I show a functional dissection of mouse gene and phenotype networks and investigate the potential that ontology-compliant phenotype annotations can offer for functional classification of genes. The mouse genome and phenome show modularity at higher levels of cellular, physiological and organismal function. Using high-throughput protein-protein interaction data, the mouse proteome was dissected and computationally extracted communities were used to predict phenotypes of mouse gene ablation. Precision and recall curves show comparable performance for higher levels of the MP ontology to those undertaken by comprehensive mouse gene function prediction such as the Mouse Function Project which predicted Gene Ontology terms. I also developed and tested an automatic procedure that relates mouse phenotypes to human diseases and demonstrate its application to the use cases of identifying mouse models given a query consisting of a set of mouse phenotypes and breaking down human diseases into mouse phenotypes. Taken together, my results may be useful as a map for candidate gene discovery, finding how mouse networks relate to human networks and investigating the evolutionary origins of their components at higher levels of gene function.

The sociotechnical system of print-on-paper for the dissemination of textual information prevails despite widespread concerns about its sustainability. On the basis of sociotechnical transitions theory the print-on-paper system is perceived as a regime. Information technology is identified as one of the generic technologies that has the potential to address the unsustainability of the incumbent regime. Its potential effects are examined through the development of the electronic book, which is defined as those IT applications providing an alternative form of textual display to printed paper. Yet, such applications have remained marginal. According to sociotechnical transitions theory the electronic book can be seen as a niche in relation to the print-on-paper regime. An alternative conceptualisation of transitions as a process of network reconfiguration is suggested. On that basis, the electronic book is depicted as a number of emergent innovation networks. Social Network Analysis methods informed by network approaches to innovation theory are applied to visualise and discuss these emergent networks. In one of the representations, the electronic book is mapped as a sociotechnical network including organisations, users and technologies. It emerges that network formation often transgresses a distinct niche-regime divide. Patterns of network interaction are explored and assessed as to whether they represent a sociotechnical transition in progress. The analysis reveals different patterns of network formation which are indicative of prospective sociotechnical trajectories where different concepts of the electronic book are emphasised. It emerges that the discussion of sustainability and the emergence of the electronic book remain largely unlinked.

With the ever increasing amount of high-throughput molecular profile data, biologists need versatile tools to enable them to quickly and succinctly analyze their data. Furthermore, pathway databases have grown increasingly robust with the KEGG database at the forefront. Previous tools have color-coded the genes on different pathways using differential expression analysis. Unfortunately, they do not adequately capture the relationships of the genes amongst one another. Structure Enrichment Analysis (SEA) thus seeks to take biological analysis to the next level. SEA accomplishes this goal by highlighting for users the sub-pathways of a biological pathways that best correspond to their molecular profile data in an easy to use GUI interface.

This thesis aims to analyze the situation on the field of public business and social networks in the world and Czech Republic. It represents the types of software designed for implementation in a business environment and subsequent processing of the relevant data obtained from these networks. The second part focuses on the area of the Public Relations in the local business environment, identifying opportunities for evaluating the contribution for the company and discusses how PR links with emerging trends of everyday use of virtual social networks in companies and in corporate environments.