Digitized by Kansas Correctional Industries

Mo, Mingzhen. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2011. / Includes bibliographical references (p. 111-123). / Abstracts in English and Chinese. / Abstract --- p.i / Acknowledgement --- p.v / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Overview --- p.1 / Chapter 1.2 --- Thesis Contributions --- p.5 / Chapter 1.3 --- Problem Description --- p.6 / Chapter 1.4 --- Thesis Organization --- p.8 / Chapter 2 --- Background Study --- p.11 / Chapter 2.1 --- Overview --- p.11 / Chapter 2.2 --- Problem Definitions --- p.12 / Chapter 2.3 --- Privacy in Online Social Networks --- p.14 / Chapter 2.4 --- Attack --- p.17 / Chapter 2.4.1 --- Statistical Learning --- p.18 / Chapter 2.4.2 --- Graph Theory --- p.22 / Chapter 2.5 --- Protection --- p.23 / Chapter 2.5.1 --- Clustering-Based Approach --- p.24 / Chapter 2.5.2 --- Modification-Based Approach --- p.27 / Chapter 3 --- Exploit Social Networks with SSL --- p.30 / Chapter 3.1 --- Overview --- p.31 / Chapter 3.2 --- Semi-Supervised Learning Framework --- p.35 / Chapter 3.2.1 --- Co-Training SSL --- p.36 / Chapter 3.2.2 --- Graph-Based SSL --- p.38 / Chapter 3.2.3 --- Local and Global Consistency Graph-Based SSL --- p.39 / Chapter 3.3 --- Experiment --- p.40 / Chapter 3.3.1 --- Dataset Description --- p.41 / Chapter 3.3.2 --- Data Preprocessing --- p.43 / Chapter 3.3.3 --- Experiment Process --- p.45 / Chapter 3.3.4 --- Experiment Results --- p.47 / Chapter 3.4 --- Conclusion --- p.49 / Chapter 4 --- Exploiting Social Networks with CG SSL --- p.50 / Chapter 4.1 --- Overview --- p.51 / Chapter 4.2 --- Exploit Learning Model and Algorithms --- p.56 / Chapter 4.2.1 --- Exploit Learning Model --- p.57 / Chapter 4.2.2 --- Algorithms --- p.60 / Chapter 4.2.3 --- Community Generation --- p.65 / Chapter 4.3 --- Experiment --- p.66 / Chapter 4.3.1 --- Dataset Description --- p.67 / Chapter 4.3.2 --- Data Preprocessing --- p.70 / Chapter 4.3.3 --- Experiment Process --- p.72 / Chapter 4.3.4 --- Experiment Results --- p.77 / Chapter 4.4 --- Conclusion --- p.82 / Chapter 5 --- APA Comparison Scheme --- p.83 / Chapter 5.1 --- Overview --- p.84 / Chapter 5.2 --- Attack-Protect-Attack (APA) Comparisons Scheme --- p.87 / Chapter 5.2.1 --- Algorithm --- p.87 / Chapter 5.2.2 --- Attack & Protection Approaches --- p.88 / Chapter 5.3 --- Experiment --- p.91 / Chapter 5.3.1 --- Dataset Description --- p.92 / Chapter 5.3.2 --- Data Preprocessing --- p.92 / Chapter 5.3.3 --- Experiment Process --- p.94 / Chapter 5.3.4 --- Experiment Result --- p.95 / Chapter 5.4 --- Conclusion --- p.103 / Chapter 6 --- Conclusion and Future Work --- p.105 / Chapter 6.1 --- Conclusion --- p.105 / Chapter 6.2 --- Future Work --- p.107 / Bibliography --- p.111

Empirical evidence shows that ensembles with adequate levels of pairwise diversity among a set of accurate member algorithms significantly outperform any of the individual algorithms. As a result, several diversity measures have been developed for use in optimizing ensembles. We show that diversity measures that properly combine the diversity space in an additive and multiplicative manner, not only result in ensembles whose accuracy is comparable to the naive ensemble of choosing the most accurate learners, but also results in ensembles that are significantly more efficient than such naive ensembles. In addition to diversity measures found in the literature, we submit two measures of diversity that span the diversity space in unique ways. Each of these measures considers not only the diversity of ratings between a pair of algorithms, but how this diversity relates to the target values.

This thesis focuses on measuring extreme risks in insurance business. We mainly use extreme value theory to develop asymptotics for risk measures. We also study the characterization of upper comonotonicity for multiple extreme risks. Firstly, we conduct asymptotics for the Haezendonck--Goovaerts (HG) risk measure of extreme risks at high confidence levels, which serves as an alternative way to statistical simulations. We split the study of this problem into two steps. In the first step, we concentrate on the HG risk measure with a power Young function, which yields certain explicitness. Then we derive asymptotics for a risk variable with a distribution function that belongs to one of the three max-domains of attraction separately. We extend our asymptotic study to the HG risk measure with a general Young function in the second step. We study this problem using different approaches and overcome a lot of technical difficulties. The risk variable is assumed to follow a distribution function that belongs to the max-domain of attraction of the generalized extreme value distribution and we show a unified proof for all three max-domains of attraction. Secondly, we study the first- and second-order asymptotics for the tail distortion risk measure of extreme risks. Similarly as in the first part, we develop the first-order asymptotics for the tail distortion risk measure of a risk variable that follows a distribution function belonging to the max-domain of attraction of the generalized extreme value distribution. In order to improve the accuracy of the first-order asymptotics, we further develop the second-order asymptotics for the tail distortion risk measure. Numerical examples are carried out to show the accuracy of both asymptotics and the great improvements of the second-order asymptotics. Lastly, we characterize the upper comonotonicity via tail convex order. For any given marginal distributions, a maximal random vector with respect to tail convex order is proved to be upper comonotonic under suitable conditions. As an application, we consider the computation of the HG risk measure of the sum of upper comonotonic random variables with exponential marginal distributions. The methodology developed in this thesis is expected to work with the same efficiency for generalized quantiles (such as expectile, Lp-quantiles, ML-quantiles and Orlicz quantiles), quantile based risk measures or risk measures which focus on the tail areas, and also work well on capital allocation problems.

Random Forests (RF) (Breiman 2001; Breiman and Cutler 2004) is a completely nonparametric statistical learning procedure that may be used for regression analysis and. A feature of RF that is drawing a lot of attention is the novel algorithm that is used to evaluate the relative importance of the predictor/explanatory variables. Other machine learning algorithms for regression and classification, such as support vector machines and artificial neural networks (Hastie et al. 2009), exhibit high predictive accuracy but provide little insight into predictive power of individual variables. In contrast, the permutation algorithm of RF has already established a track record for identification of important predictors (Huang et al. 2005; Cutler et al. 2007; Archer and Kimes 2008). Recently, however, some authors (Nicodemus and Shugart 2007; Strobl et al. 2007, 2008) have shown that the presence of categorical variables with many categories (Strobl et al. 2007) or high colinearity give unduly large variable importance using the standard RF permutation algorithm (Strobl et al. 2008). This work creates simulations from multiple linear regression models with small numbers of variables to understand the issues raised by Strobl et al. (2008) regarding shortcomings of the original RF variable importance algorithm and the alternatives implemented in conditional forests (Strobl et al. 2008). In addition this paper will look at the dependence of RF variable importance values on user-defined parameters.

No description available.

This study arose out of a search for a suitable scale to measure religion and spirituality. The literature suggests that religion and spirituality are potentially powerful explanatory variables in health and social research, but there do not appear to be any instruments that are generally accepted as measures of an individual's religious or spiritual characteristics. While a lack of consensus in such a complex area is probably to be expected, it is the lack of accepted measures or instruments that drives this study. The literature review describes the historical influence of religion on public health practices, and the most recently reported associations between religion and both physical and mental health. This establishes religion as a potentially useful construct to include in any health study. However, the reported association between religion and health is often unclear, and the measures used differ widely between studies. This study goes beyond the health context and explores the reasons why existing methods have not resulted in broadly accepted measures of religion and spirituality.

This thesis proposes new algorithms for restoration strategies that provision bandwidth guaranteed recovery for unicast and multicast connections. The primary focus is on online restoration strategies that sequentially do pre-planning of resource for each request using the current network resource state. Online restoration strategies do not require prior knowledge of all the requests like that of offline restoration strategies. Therefore, online restoration strategies are more suitable for on-demand and dynamic traffic engineering control. The proposed new algorithms are compared to known algorithms from literature. Most literature evaluates the performance of the algorithms with two metrics only: total bandwidth requirement and the number of requests accepted in the network. This thesis evaluates the algorithms in one additional dimension: the computational time. This is an important criterion when response times for establishing new connections are stringent. Each algorithm makes trade-off between computational complexity, bandwidth efficiency, and number of accepted requests. Results show that the proposed algorithms provide alternative trade-offs between the three performance metrics when compared to other existing algorithms. The alternatives provide more choice for the network providers and the best algorithm to use depends on the network's requirements. The restoration strategies used for unicast and multicast connections in this thesis are very compatible thus it is possible to integrate the restoration strategies into a single system where they share the same backup resources. Results from simulations show that using an integrated restoration model has significant benefits, which includes lower backup bandwidth requirement than the separate restoration model.

Many types of foods are sterilised and packaged aseptically to maximise their shelf life and ensure that they are safe for consumption. However, if even a few micro-organisms survive these treatments, the quality and/or safety of the food can be seriously compromised. Therefore, packaging materials that can actively participate in the process of inactivating micro-organisms have a potentially important role in protecting the quality and safety of packaged foods. The aim of this project is to explore the effectiveness of the photodynamic effect as a method of inactivating micro-organisms on the surface of packaging materials. Photodynamic action occurs when a photosensitiser molecule absorbs light in the presence of oxygen, leading to the formation of reactive oxygen species such as singlet oxygen, superoxide radicals, or other free radicals; these species are known to inactivate cells. Medical uses of photodynamic action have been studied extensively, usually in applications where the reactive oxygen species are produced intracellularly. However, very little research has been conducted to investigate the efficacy of extra-cellularly generated reactive oxygen species on the viability of micro-organisms, and accordingly this research project investigated the possibility that this approach might be used for inactivating micro-organisms on packaging materials. The results of this study indicate that singlet oxygen may be produced at the surface of polymer films that contain selected photosensitisers, resulting in the oxidation of known singlet-oxygen acceptors to form endoperoxides. This provides evidence that singlet-oxygen mediated reactions can occur at the surface of a sensitising film. It is likely that cell destruction is caused by singlet-oxygen, rather than involving other activated species. It was further shown that the photodynamically generated singlet-oxygen can inactivate micro-organisms (extra-cellularly) on the surface of a polymeric material. The study included examples from each genus of micro-organisms that are of concern to the food and packaging industry, including: Gram-negative bacteria (Escherichia coli); Gram-positive bacteria (Bacillus cereus), both vegetative cells and endospores; yeast (Saccharomyces cerevisiae), and mould (Fusarium oxysporium.) Results indicate that the photodynamic effect causes a substantial reduction in viable cell numbers for vegetative cells and spores (both bacterial and fungal) that have been inoculated onto a plastic surface containing the photosensitiser, anthraquinone,. The results show that an increase in the amount of reactive oxygen species produced by photodynamic action increases the inactivation rate of the micro-organisms. The micro-organisms investigated were susceptible to photodynamic action to varying extents. In conclusion, the results demonstrate that control of microbial populations on the surface of polymeric films (used in food packaging) is achievable using photodynamic action produced from polymers, based on the reported amounts of micro-organisms found on food packaging materials. / Master of Science (Hons)

The need for information anywhere and at any time has been the driving force for the increasing growth in mobile networks and devices. The field of mobile computing is the merger of advances in computing and communications with the aim of providing seamless and ubiquitous computing environment for mobile users. Whereas notebook computers and personal digital assistants (PDAs) are self-contained, networked computing constitutes a new paradigm of computing that is revolutionizing the way computers are used. Mobile networking greatly enhances the utility of carrying a computing device. It provides mobile users with versatile communication to other people and expedient notification of important events, yet with much more flexibility than cellular telephones and pagers. It also permits continuous access to services and resources of the traditional land-based wired networks. This combination of networking and mobility will engender new applications and services, such as collaborative software to support impromptu meetings, electronic bulletin boards that adapt to the contents according to the participants present, self adjusting lighting and heating, and navigation software to guide users in unfamiliar places and tours. To support mobility in the Internet, the Internet Protocol (IP) has been extended to support mobility. Also at the same time, there is also a growing trend for these IP based networks to operate in an infrastructureless environment called mobile ad-hoc networks. However, the proliferation of such mobile networks depends on a multitude of factors, with trustworthiness being one of the primary challenges to be met. The objective of this dissertation is to address the issues involved in the design of security services for Mobile IP and ad-hoc networks. Extensions to IP based networks (both wired and infrastructureless networks) to facilitate mobility have not been designed keeping security in mind. However adequate security features are basic requirements for the continued functioning of mobile networks. Clearly the problem is so broad that there is no way to devise a general solution We aim to address most of these wide- ranging problems and in the process initiate a practical approach to the development of an integrated security infrastructure for mobile networks. The intention is to seamlessly integrate these security services and mechanisms at the IP level within the mobile IP and ad-hoc networks. The provision of security services at the higher and lower layers and their interoperability with our proposed framework is outside the scope of this thesis / Doctor of Philosophy (PhD)