<p>CHANG, JIAE. Development of a perceptual Visualization assitant for multi-dimensionaldata visualizations. (Under the direction of Christopher G. Healey.)This thesis describes a perceptual visualization assistant (ViA) designed to help usersconstruct perceptually salient visualizations to display high dimensional data stored in amulti-dimensional dataset. ViA focuses on (1) how to construct visualizations in an application-independent manner, (2) how to produce visualizations that support effectiveexploration, discovery, and analysis, and (3) how to allow users to participate in this processto guarantee the resulting visualizations respect context and user preferences. Tosolve these issues we propose guidelines based on studies of human vision and human perception.Evaluation engines built from these guidelines will be used to evaluate the dataattribute to visual feature mappings in a given visualization. The results (evaluationweights and hints for improvement) will be used by a mixed-initiative search algorithm toidentify new visualizations with high potential for improvement. This allows ViA to selectivelysearch for the best possible visualizations for a user?s dataset and analysis tasks.ViA allows user interaction during the search process by asking whether certain simplifyingassumptions can be made, and by allowing users to specify requirements and constraintsfor the visualizations ViA suggests. We tested ViA?s abilities in two practicalapplication environments: weather data and e-commerce auction traces. Results from asubjective user questionnaire suggest that ViA?s visualizations are as good or better thanexisting standards used to represent this type of data.<P>

<p>AbstractWeili, Zhu. The Application of Monte Carlo Sampling to Sequential Auction Games with Incomplete Information: -An Empirical Study. (Under the direction of Peter Wurman.)In this thesis, I develop a sequential auction model and design a bidding agent for it. This agent uses Monte Carlo sampling to ¡°learn¡± from a series sampled games. I use a game theory research toolset called GAMBIT to implement the model and collect some experimental data. The data shows the effect of different factors that impact on our agent¡¯s performance, such as the sample size, the depth of game tree, etc. The data also shows that our agent performs well compared with myopic strategic agent. I also discuss the possible relaxation of different aspects in our auction model, and future research directions. <P>

<p>GARLICK, JOANNA SHANNON. The Inclusion of Reference Seeds in Generated Dialogue (Under the direction of Michael Young).<P>In the course of natural dialogue, humans are able to effectively use previous conversation. This is true for many types of dialogue, including instructional or task-related conversation. When human speakers effectively leverage past conversation for illustrative or comparative purposes, their hearers reap the benefits of increased dialogue coherency and are better able to assimilate knowledge. The practice of utilizing past utterances in current conversation is called referencing. It is believed that if computer agents could utilize referencing the way humans do, these agents would produce more natural-sounding and effective dialogue. </P><P>For this reason, the natural language community has conducted a fair amount of research into referencing. While most of the past and current research focuses on identifying opportunities to refer backward and how to do so, this research focuses on the novel concept of a reference seed. We introduce the term seed to refer to a piece of information opportunistically included in generated dialogue with the intent that it be referred back to in the future. We show how seeds and references can be implemented using discourse planning operators and identify some general rules for when to do so. We also discuss a test system implemented to showcase the potential of the concept and report the encouraging results of an informal user study we have conducted. </P><P>

<p>Combinatorial auctions are a promising way of solving the complicated heterogeneous resource allocation problem. However combinatorial auctions present inherently hard computational problems for agents and auctioneers. In particular (1) the auction must solve the winner determination problem, which is NP-complete and (2) the agents must solve the value determination problem which can be intractable when computing a precise valuation for the bundles.The winner determination problem can be addressed by (1) developing algorithms which solve the problem of computing the optimal solution; (2) developing heuristics which approximate the solution quickly. The combinatorial auctions, proposed in the literature are dependent on the optimal solution of the winner determination problem, which is intractable for large number of agents and items.In this thesis we study the effects of using approximate winner determination in a particular iterative combinatorial auction : A1BA. We categorize the effects of approximation and develop several corrective methods which reduce the errors due to approximation in several categories.<P>

<p>Numerous studies have shown that interpersonal communication acts as animportant channel for gathering information. But if we wish to rely oninterpersonal communication, we still need to figure out how to determinethe right person to ask. Usually we cannot find the potential expert(s)directly, and we need some assistance from our friends or friends'friends to locate them. The phenomenon of Six Degrees of Separationindicates that it is possible to use some intelligent software agents,who can interpret the links between people and follow only therelevant one, to find the desired experts quickly. A computational model of agent-based referral networks was proposed to assistand simplify the users to find potential experts for a specified topic in aperson-to-person social network, in which each user is assigned a softwareagent, and software agents help automate the process by a series of ``referralchains''. Unlike most previous approaches, our architecture is fullydistributed and includes agents who preserve the privacy and autonomy oftheir users. These agents learn models of each other in terms of expertise(ability to produce correct domain answers), and sociability (ability toproduce accurate referrals). We study this framework experimentally to seethe effects that the different variables have on each other and the efficiencyof the referral networks. Furthermore, a social mechanism of reputation management was proposed tohelp agents (users) avoid interaction with undesirable participants inthe referral networks. The mathematical theory of evidence is used torepresent and propagate the reputation information in a referral network.Our approach adjusts the ratings of agents based on their observations aswell the testimony from others. Moreover, we conducted several experimentsto study the reputation management in different settings. Social mechanismsare even more important when some centralized reputation managementmechanisms, i.e., trusted third parties, are not available. Our specificapproach to reputation management leads to a decentralized society in whichagents help each other weed out undesirable players.<P>

<p><P>As networks become increasingly more open and complex, their management becomes that much more important. Numerous commercial and open-source tools already exist that are capable of providing useful network analysis. Unfortunately, the majority of these commercial tools are either quite expensive, or require a significant amount of effort for their deployment. Other tools are very specific in their tasks, and offer little in the way of customization. Tools able to provide favorable statistics often fall short when it comes to operating on loaded high-bandwidth networks. Also, as the majority of network management tools operate at the packet level, they often require administrator-level access to capture the data. In addition, there are privacy issues that may also limit who has access to what part of the data. On large networks, this can amount to a lot of data that needs to be scrutinized by a limited number of people.</P><P>The purpose of this project was to develop an inexpensive, customizable network-monitoring tool (called Resource Usage Monitor) that a) is capable of providing a variety of traffic-related statistical data on a high-bandwidth network, b) provides a user-friendly selective access to that data for users with different privacy privileges, and c) interfaces to a policy management toolset to allow pro-active management of the network based on security, quality and resource information it gathers. The processing engine behind the Resource Usage Monitor (RUM) examines data at the monitored gateway, collecting inbound and outbound information on the number of bytes, packets, network and application flows, and such, transmitted or received for each internal and external host. A web interface provides persistence graphs and reports that can disclose general and specific traffic patterns on the network. This information can be used to assess security, resource usage, and quality of service (QoS) assessment of the monitored network and hosts. For example, setting flow, load, and other activity thresholds at different levels of granularity allows for the detection of anomalies throughout the network. Port scans, Denial of Service (DOS) attacks, and Trojan applications have been detected through surveillance of simple threshold-based patterns. More complex, possibly multi-probe, patterns can reveal much more subtle anomalies and side effects.</P><P>RUM operates in a statistical mode, rather than continuous mode. It samples the network every few minutes. After each sample the collected data is analyzed and appropriate warnings and interactions with the policy services are effected. Collected packets are sorted and stored by pre-defined subnets allowing parallelism in the processing of the data. This also separates datasets for another reason; it enables secure access to just the network traffic for which an administrator is responsible. Persistence data is kept in the form of graphs generated using RRDtool, a round robin database utility, and in logs. The information logged is completely customizable and can even be offloaded for analysis by other systems. This thesis describes the RUM architecture, data it collects, analysis modules, user handling, and other features.</P><P>

<p>Knowledge Management has been the subject of increasing focus over the last several years. Literature and research on the topic has grown as companies and organizations have come to realize that success is often determined by one's ability to create, disseminate, and embody knowledge in products and services. This realization has led to increased interest in examining the ways in which knowledge can be effectively created, identified, codified, disseminated, and retained. The field of Knowledge Management has emerged to address this need.<p>One of the obstacles that Knowledge Management seeks to overcome is the natural tendency in people to hoard knowledge. People often withhold knowledge when they feel it provides them with a competitive advantage over others. Many traditional management incentives and team structures create and perpetuate competitive environments that encourage knowledge hoarding. Knowledge Management also seeks to find ways to reduce the impact of employee turnover. When an employee leaves a company or organization, the knowledge they possess often goes with them. This loss can potentially have a negative impact on the productivity and quality of the company or organization. Knowledge Management seeks to find ways to minimize loss of knowledge when an employee leaves a company or organization.<p>Pair programming is a practice that holds promise for overcoming some of the challenges faced by Knowledge Management. In pair programming, two programmers work side-by-side at one computer collaborating on the same design, algorithm, code, or test. The continual interaction between pair programmers would seem to provide an environment that promotes knowledge sharing, and collaborative knowledge discovery. Additionally, through pair rotation, in which pairs change partners fairly often, tacit knowledge might be spread more effectively through face-to-face communication than by documentation, databases, or other means.<p>This research examines pair programming in the realm of Knowledge Management, positing the following hypotheses:Pair programming reduces the tendency of people to hoard knowledge.Pair programming reduces the impact of employee turnover.Pair programming is an effective means of knowledge dissemination and knowledge retention that has a positive influence on the Knowledge Management practices of a company or organization.</ul><p>These hypotheses are tested through the use of a survey of individuals in technology research, development, and service. Analysis of the survey results provided no conclusive evidence to either support or disprove the hypothesis that pair programming reduces the tendency of people to hoard knowledge. The results indicate support for the hypothesis that pair programming reduces the impact of employee turnover, although not statistically significant. And finally, the survey results indicate with statistical significance that pair programming is an effective means of knowledge dissemination and retention, with a positive influence on the Knowledge Management practices of a company or organization. <P>

by<p>Daniel C. DuVarney<p>PhD Dissertation submitted to the Faculty of the North Carolina State University in partial fulfillment of the requirements for the degree of <p>PhD<P>in<P>Computer Science<P>Approved<br><br>S. Purushothaman Iyer, Chair<br>W. Rance Cleaveland<br>K.C. Tai<br>John W. Baugh<p>January 22, 2002<br>Raleigh, North Carolina<p>Abstract<p>Model checking is a major advancement in the quest forpractical automatic verification methods for computer systems, and has been effectively used to discoverflaws in real-world hardware systems. Unfortunately, applying model-checking techniques to software systems has proved to be more difficult, due to the large number of states and irregular transitions of such systems. One promising method for generating reasonably-sized models from programs is the use of data abstraction, in which the program data is mapped from a large set of possible values to a much smaller set of abstract values. This thesis develops a method which, given a program in the C language and an abstraction mapping, allows the automatic construction of an abstract labeled transition system (LTS), which is much smaller than the concrete LTS (the LTS which would be generated without the benefit of abstraction). The method is shown to be sound in the sense that if a program is well-behaved in its use of pointers, then any linear temporal logic formula which holds true for the corresponding abstract LTS will also hold true for the concrete LTS. Furthermore, if a design exists in the form of a transition system, then the abstract LTS can be checked against the design for bisimilarity. Bisimilarity ensures that the program is a faithful implementation ofthe design. A suite of software tools has been implemented based upon the theory. These tools interface with the Concurrency Workbench, a model checking system. A case study is presented which shows the practicality of this technique for verifying real-world C programs.<P>

<p>People waste a lot of time waiting in queues. Even when they go to places where they can make reservations, the reservations alone don?t provide much flexibility. The purpose of this research is to show that the use of mobile devices, which allow us to improve planning, trade reservations and receive total information about queues, improves the social welfare. Several experiments were performed using a custom made simulator. The results of the experiments demonstrate that a simple reservation mechanism can even reduce the social welfare under certain conditions, but tradable reservations and clairvoyance each improve it. While in many situations queues are unavoidable, better information and more flexibility in reservation handling can improve the overall quality of societies use of these resources.<P>

Thesis (M.S.) -- University of Texas at Dallas, 2006. / Includes vita. Includes bibliographical references (leaves 174-176)