Thesis (M.A.(Ed.))--University of Hong Kong. / Also available in print.

Thesis (Ed.D.) -- Central Connecticut State University, 2005. / Thesis advisor: Richard Arends. "... in partial fulfillment of the requirements for the degree of Doctor of Education." Includes bibliographical references (leaves 61-72). Also available via the World Wide Web.

Thesis (LL. M.)--Judge Advocate General's School, U.S. Army, 1968. / "April, 1968." Typescript. Includes bibliographical references (leaves 71-72). Also issued in microfiche.

It is 2am. A professor wakes up with a new direction for her research; she must immediately learn about bioethics. In a dorm a student is finally ready to begin a paper on Cuba. Where do they turn? The library web site presents them with a bewildering array of resources and no librarian on hand to serve as intermediary. How can librarians facilitate research in their absence? What interfaces can be designed to educate users in their search? What metadata is needed to enable accurate retrieval? What is the librarian’s role in the increasingly indirectly-mediated information-seeking environment? Can the reference interview be effectively translated into a search interface? This paper describes a step towards resolving these issues by creating an on-line tool to assist users in selecting the database(s) most germane to their research needs.

A partir de la experiencia del autor como guionista en la serie “Puertas al más allá”, se hace un análisis del trabajo profesional del guion en una serie de televisión producida para el exterior.

Due to the combination of constrained power, low duty cycle, and high mobility, neighbor discovery is one of the most challenging problems in wireless sensor networks. Existing discovery designs can be divided into two types: pairwise-based and group-based. The former schemes suffer from high discovery delay, while the latter ones accelerate the discovery process but increase transmission package size or incur too much energy overhead, far from practical. Guided by the Talk More Listen Less (TMLL) principle (published in 2016), in which beacons are not necessarily placed in the wakeup slots, we propose two different versions of a group-based protocol we called Talk Half Listen Half (THLH). For the first time, a group-based protocol uses the Channel Occupancy Rate (COR), one of the fundamental novel components of the TMLL model, for performance improvements, in the same way, Duty Cycle (DC) was used in previous group-based protocols. Both versions of the protocol use low transmission overhead in comparison with previous group-based discoveries. After analyzing pros and cons of each approach, we arrived at the conclusion that both behave the best for networks where the average number of new neighbors per slot (β) is low, a metric that sets the bases for performance comparisons of any current/future work with variable COR usage. We also derived a formula that links this new metric with the worst case avg. COR usage of our proposed protocols. Finally, simulation results show that our protocol can improve the average discovery latency and worst case latency close to 50% given low β values.

Grid computing is a virtualized distributed computing environment aimed at enabling the sharing of geographically distributed resources. Grid resources have traditionally consisted of dedicated supercomputers, clusters, or storage units. With the present ubiquitous network connections and the growing computational and storage capabilities of modem everyday-use computers, more resources such as PCs, devices (e.g., PDAs and sensors), applications, and services are on grid networks. Grid is expected to evolve from a computing and data management facility to a pervasive, world-wide resource-sharing infrastructure. To fully utilize the wide range of grid resources, effective resource discovery mechanisms are required. However, resource discovery in a global-scale grid is challenging due to the considerable diversity, large number, dynamic behavior, and geographical distribution of the resources. The resource discovery technology required to achieve the ambitious global grid vision is still in its infancy, and existing applications have difficulties in achieving both rich searchability and good scalability. In this thesis, we investigate the resource discovery problem for open-networked global-scale grids. In particular, we propose a distributed semantics-based discovery framework. We show how this framework can be used to address the discovery problem in such grids and improve three aspects of performance: expressiveness, scalability, and efficiency. Expressiveness is the first characteristic that a grid resource-searching mechanism should have. Most existing search systems use simple keyword-based lookups, which limit the searchability of the system. Our framework improves search expressiveness from two directions: First, it uses a semantic metadata scheme to provide users with a rich and flexible representation mechanism, to enable effective descriptions of desired resource properties and query requirements. Second, we employ ontological domain knowledge to assist in the search process. The system is thus able to understand the semantics of query requests according to their meanings in a specific domain; this procedure helps the system to locate only semantically related results. The more expressive the resource description and query request, however, the more difficult it is to design a scalable and efficient search mechanism. We ensure scalability by reconfiguring the network with respect to shared ontologies. This reconfiguration partitions the large unorganized search space into multiple well-organized semantically related sub-spaces that we call semantic virtual organizations. Semantic virtual organizations help to discriminatively distribute resource information and queries to related nodes, thus reducing the search space and improving scalability. To further improve the efficiency of searching the virtual organizations, we propose two semantics-based resource-integrating and searching systems: GONID and OntoSum. These two systems address searching problems for applications based on different network topologies: structured and unstructured peer-to-peer overlay networks. Queries in the search systems are processed in a transparent way, so that users accessing the data can be insulated from the fact that the information is distributed across different sources and represented with different formats. In both systems, ontological knowledge is decomposed into different coarse-grained elements, and then these elements are indexed with different schemes to fit the requirements of different applications. Resource metadata reasoning, integrating, and searching are based on the index. A complex query can be evaluated by performing relational operations such as select, project, and join on combinations of the indexing elements. We evaluate the performance of our system with extensive simulation experiments, the results of which confirm the effectiveness of the design. In addition, we implement a prototype that incorporates our ontology-based virtual organization formation and semantics-based query mechanisms. Our deployment of the prototype verifies the system's feasibility and its applicability to real-world applications. / Science, Faculty of / Computer Science, Department of / Graduate

Philosophiae Doctor - PhD / There are over 100 different types of cancer, and each of these cancers are classified by the type of cell that it initially affects. For the purpose of this research we will be focussing on prostate cancer (PC). Prostate cancer is the second most common form of cancer in men around the world and annually approximately 4500 men in South Africa are diagnosed making PC a global epidemic. Prostate cancer is a type of cancer which starts in the prostate it is normally a walnut-sized gland found right below the bladder. PC follows a natural course, starting as a tiny group of cancer cells that can grow into a tumour. In some men if PC is not treated it may spread to surrounding tissue by a process called direct invasion/ spread and could lead to death. Current diagnostic tests for prostate cancer have low specificity and poor sensitivity. Although many PC's are slow growing there is currently no test to distinguish between these and cancers that will become aggressive and life threatening. Therefore the need for a less invasive early detection method with the ability to overcome the lack of specificity and sensitivity of current available diagnostic test is required. Biomarkers have recently been identified as a viable option for early detection of disease for example biological indicators ie. DNA, RNA, proteins and microRNAs (miRNAs). Since first described in the 1990s, circulating miRNAs have provided an active and rapidly evolving area of research that has the potential to transform cancer diagnostics and prognostics. In particular, miRNAs could provide potentially new biomarkers for PC as diagnostic molecules. Circulating miRNAs are highly stable and are both detectable and quantifiable in a range of accessible bio-fluids, having the potential to be useful as diagnostic, prognostic and predictive biomarkers. In this study we aimed to identify miRNAs as potential biomarkers to detect and distinguish between various types of PC in its earliest stage. The major objectives of the study were to identify miRNAs and their gene targets that play a critical role in disease onset and progression to further understand their mechanism of action in PC using several in silico methods, and to validate the potential diagnostic miRNAs using qRT-PCR in several cell lines. The identification of specific miRNAs and their targets was done using an "in-house" designed pipeline. Bioinformatic analyses was done using a number of databases including STRING, DAVID, DIANA and mFold database, and these combined with programming and statistical analyses was used for the identification of potential miRNAs specific to PC. Our study identified 40 miRNAs associated with PC using our "in-house" parameters in comparison to the 20-30 miRNAs known to be involved in PC found in public databases e.g. miRBase. A comparison between our parameters and those used in public databases showed a higher degree of specificity for the identification PC-associated miRNAs. These selected miRNAs were analysed using different bioinformatics tools, and were confirmed to be novel miRNAs associated with PC. The identified miRNAs were experimentally validated using qRT-PCR to generate expression profiles for PC as well as various other cancers. Prostate lines utilised in this study included PNT2C2 (normal) which was compared to BPH1 (Benign) and LNCaP (Metastatic). In the study the expression profiles of eight potential miRNA biomarkers for the detection of PC was determined using qRT-PCR, and to distinguish PC from other cancers. QRT-PCR data showed that miRNA-3 and -5 were up-regulated in the BPH1 and LNCaP when compared to PNT2C2. In addition miRNA-8 was also shown to be up-regulated in LNCaP. Based on these results it was shown that a miRNA profile could be established to distinguish between BPH1 and the LNCaP prostate cell lines. The results suggest that one miRNA as a diagnostic marker may be sufficient to differentiate between different cancer cell lines. Furthermore by creating a unique profile for each cancer cell line by using a combination of miRNAs could be a suitable approach as well. Finally, it was shown that through the use of a single or combination of all eight miRNAs a unique profile for all the cancer cell lines tested in this study can be created. This is an important finding which could have potential diagnostic or prognostic implications in clinical practice.

Pattern Discovery, a popular paradigm in data mining refers to a class of techniques that try and extract some unknown or interesting patterns from data. The work carried out in this thesis concerns frequent episode mining, a popular framework within pattern discovery, with applications in alarm management, fault analysis, network reconstruction etc. The data here is in the form of a single longtime-ordered stream of events. The pattern of interest here, namely episode, is basically a set of event-types with a partial order on it. The task here is to unearth all patterns( episodes here) which have a frequency above a user-defined threshold irrespective of pattern size. Most current discovery algorithms employ a level-wise a priori-based method for mining, which basically adopts a breadth-first search strategy of the space of all episodes. The episode literature has seen multiple ways of defining frequency with each definition having its own set of merits and demerits. The main reason for different frequencies definitions being proposed is that, in general, counting all occurrences of a set of episodes is computationally very expensive. The first part of the thesis gives a unified view of all the apriori-based discovery algorithms for serial episodes(associated with a total order)under these various frequencies. Specifically, the various existing counting algorithms can be viewed as minor modifications of each other. We also provide some novel proofs of correctness for some of the serial episode counting schemes, which in turn can be generalized to episodes with general partial orders. Our unified view helps us derive quantitative relationships between different frequencies. We also discuss all the anti-monotonicity properties satisfied by the various frequencies, a crucial information needed for the candidate generation step. The second part of the thesis proposes discovery algorithms for episodes with general partial orders, for which no algorithms currently exist in literature. The discovery algorithm proposed is apriori-based and generalizes the existing serial and parallel (associated with a trivial order) episode algorithms. The discovery algorithm is a level-wise procedure involving the steps of candidate generation and counting a teach level. In the context of general partial orders, a major problem in a priori-based discovery is to have an efficient candidate generation scheme. We present a novel candidate generation algorithm for mining episodes with general partial orders. The counting algorithm design for general partial order episodes draws ideas from the unified view of counting for serial episodes, presented in the first part of the work. We formally show the correctness of the proposed candidate generation and counting steps for general partial orders. The proposed candidate generation algorithm is flexible enough to be able to mine in certain specialized classes of partial orders (satisfying what we call maximal sub episode property), of which, the serial and parallel class of episodes are two specific instances. Our algorithm design initially restricts itself to the class of general partial order episodes called injective episodes wherein repeated event-types are not allowed. We then generalize this to a larger class of episodes called chain episodes, where episodes can have some repeated event types. The class of chain episodes contains all (including non-injective) serial and parallel episodes and thus our method properly generalizes the existing methods for serial and parallel episode discovery. We also discuss some problems in extending our algorithms to episodes beyond the class of chain episodes. Also, we demonstrate that frequency alone is not a sufficient enough interestingness measure for episodes with unrestricted partial orders. To address this issue, we propose an additional measure called bidirectional evidence to assess interestingness which, along with frequency is found to be extremely effective in unearthing interesting patterns. In the frequent episode framework, the choice of thresholds are most often user-defined and arbitrary. To address this issue, the last part of the work deals with assessing significance of partial order episodes in a statistical sense based on ideas from classical hypothesis testing. We declare an episode to be significant if its observed frequency in the data stream is large enough to be very unlikely, under a random i.i.d model .The key step in the significance analysis involves the mean and variance computation of the the time between successive occurrences of the pattern. This computation can be reformulated as, solving for the mean and variance of the first visit time to a particular stat e in an associated Markov chain. We use a generating function approach to solve for this mean and variance. Using this and a Gaussian approximation to the frequency random variable, we can now calculate a frequency threshold for any partial order episode, beyond which we infer it to be significant. Our significance analysis for general partial order episodes generalizes the existing significance analysis of serial episode patterns. We demonstrate on synthetic data the effectiveness of our significance thresholds.

During the route discovery process, each node receiving the route request packet (RReq) will retransmit it exactly once. A distant neighbor may accidentally receive/loose the only RReq and use it to announce a new route, although that link is inferior/superior for route reply packets (RRep) or actual message routing. Overall, the constructed route may be far from the optimal. All existing route discovery schemes (including DSR/AODV) apply retransmission during route discovery exactly once (1R). Based on a realistic physical layer model, we propose two new route discovery schemes: n-retransmission (nR, retransmitting exactly n times) and n-retransmission c-reception (ncRR), retransmitting until we either reach a total of n own retransmissions or c copies from neighbors are heard. We compare our two new scheme with the traditional one, under otherwise identical conditions (same metric, same packet reception probability on each link) and the same choices about possibly retransmitting again upon discovering a better route (R+) or discarding it (R1), generating route reply packet for every received RRep (B*), or for first and better discovered routes only (B2), and retransmitting RRep exactly once (A1), up to a maximum of three times (A3), or optimally u times decided by link quality (Au). Experimental results show that the proposed ncRR scheme (for n=2 and c=3 or c=4) achieves the best tradeoff between quality of route, success rate and message overhead in the route discovery process, followed by the nR scheme, and both of them are superior to the existing traditional schemes.