2007
EDGE Distinguished Lecture Series & EDGE Student Scholar Lecture Series
Thursday at 2:00 pm
ACES 2.402
(For parking, please contact Charlotte Harris, charlotte.harris@edge.utexas.edu, (512)232-3343)
 |
February 8 Dr. Don Batory Centennial Professor Computer Science |
|
From Implementation to Theory in Product Synthesis. Future software development will rely on product synthesis, i.e., the synthesis of code and non-code artifacts for a target component or application. Prior work on feature-based product synthesis can be unified by applying elementary ideas from category theory. Doing so reveals (a) important and previously unrecognized properties that product synthesis tools must satisfy, and (b) non-obvious generalizations of current techniques that will guide future research efforts in automated product development. Keynote speech given for POPL 2007: Preliminary Program. |
|
 |
September 21 Dr. Pierre J Corriveau Chief Technology Officer Naval Undersea Warfare Center |
|
Next Generation Undersea Warfare. Focus on The Grand Challenge of Distributed Network Systems. Next Gen Undersea Warfare is FULL SPECTRUM warfare emanating from and exploiting the undersea domain. We are convinced that Distributed Networked Systems will enable Next Gen USW. A Distributed Networked System is a set of sensors, unmanned vehicles, and platforms, netted and working together to provide a clear understanding, picture, and allowing for total control of the undersea battle space. We at the Naval Undersea Warfare Center will focus NUWC Science & Technology on the “Grand challenge of Distributed Networked Systems.” |
|
 |
April 27 (PLEASE NOTE: start time is 1:00pm) Sanem Kabadayi Ph.D. Student Mobile and Pervasive Computing Group Electrical and Computer Engineering |
|
Enabling Declarative Applications in Immersive Sensor Networks As sensor networks become increasingly ubiquitous, future scenarios in which multiple mobile applications will leverage sensor network nodes opportunistically and unpredictably will become possible. Such scenarios deviate from existing deployments of sensor networks which are often highly application-specific and generally funnel information to a central collection service for a single purpose. A significant barrier to the widespread development of such flexible sensor network applications lies in the increased complexity of the programming task when compared to existing distributed or even mobile situations. Sensor nodes are severely resource-constrained, in terms of both computational capabilities and battery power, and therefore an application development task must inherently consider low-level design concerns. This complexity, coupled with the increasing demand for applications, highlights the need for programming platforms (i.e., middleware) that simplify application development. This talk introduces the DAIS (Declarative Applications in Immersive Sensor networks) middleware platform that enables the development of these adaptive ubiquitous computing applications. Our approach focuses on minimizing communication in the sensor network to best ensure the network's lifetime. DAIS localizes data collection and sensor interaction to only the regions of the network required for the applications' immediate data needs. At the programming interface level, this requires exposing some aspects of the physical world to the developer, and DAIS accomplishes this through novel programming abstractions that enable on-demand access to dynamic data sources. The talk will overview the middleware and provide details of two of its components: the scene communication abstraction and virtual sensors, including both programmatic examples and performance results. |
|
 |
April 13 Christine Julien Assistant Professor Director, Mobile and Pervasive Computing Group Electrical and Computer Engineering |
|
Ubiquitous Computing Middleware: Application Sessions for Dynamic Environments Building applications for any distributed environment has always been a difficult task. Introducing the demands of mobile and pervasive devices such as dynamic connections and limited resources are further increasing these software engineering burdens. These inherent complexities are coupled with usage trends that are demanding applications be developed and deployed at alarming rates. The expertise required to rapidly build robust applications in these environments is expensive and rare. This talk will discuss the benefits of using middleware solutions to bridge the gap between the applications demanded by consumers and the capabilities of software engineers to produce those applications quickly and correctly. Middleware solutions in general abstract the complexities of an environment to simplify programming sophisticated applications. The talk will focus on a particular middleware that enables dynamic resource usage by users embedded in a pervasive computing environment. The middleware's underlying computational model demonstrates how a rigorous formalization of a middleware's operation can communicate the middleware's behavior to application developers who employ it. Finally, a set of applications that use the middleware exhibit its potential impact on real-world situations. The talk will also provide an introduction to the Mobile and Pervasive Computing Group (MPC) within EDGE and give an overview of our current projects and research results. |
|
 |
March 23 Karen K. Fullan Ph.D. Student Laboratory for Intelligent Processes and Systems Electrical and Computer Engineering |
|
Who Do You Trust? Learning Strategies for Trusting in Open Networks In open networks-or networks in which authentication requirements can be circumvented-malicious or incompetent entities may gain access, causing harm to other network members. The multi-agent system paradigm provides a unique approach to network protection by modeling network entities as agents (software acting on behalf of humans) which use trustworthiness assessment to evaluate potential network partners based on past interactions, as well as reputations communicated by other agents. Trustworthiness assessment is vital for protecting online transactions in e-commerce applications (eBay, Bizrate), isolating predators in online social networks (MySpace, orkut), and verifying information accuracy in recommendation networks (Epinions, Amazon). An agent's trust decision strategy consists of the agent's policies for making trust-related decisions, such as whom to trust, how trustworthy to be, what reputations to believe, and when to tell truthful reputations. In reputation exchange networks, learning trust decision strategies is complex; the agent's interactions with one partner may be dependent upon interactions with another since partners may tell each other about their experiences with the agent, influencing future behavior. This talk demonstrates how dependency assumptions enable agents to correlate rewards to actions, thus learning effective trust decision strategies. Experiments in the Agent Reputation and Trust (ART) Testbed, an open-source experimental marketplace, demonstrate higher earnings of strategy-learning agents over agents employing naive strategies. The variation in earnings of reputation-based learning vs. experience-based learning over different agent systems motivates dynamic utilization of reputations vs. experience in making trust decisions. |
|
 |
March 9 Sarfraz Khurshid Assistant Professor Electrical and Computer Engineering |
|
Automated Test Generation Using a Constraint Solver For programs that manipulate complex structures, generating a good test suite is difficult and laborious. This talk presents a novel approach for generating a suite completely automatically. The user needs to provide only an invariant or precondition characterizing the acceptable inputs, and a bound on input size. A tool then generates a collection of appropriate test cases, using a constraint solver as the underlying engine. The talk explains the details of this approach and describes experience applying it to various problems, ranging from library code to stand-alone applications. Experiments on checking some methods from the Java Collection Framework indicate that it is feasible to systematically generate a high quality test suite. The tool also uncovered previously unknown errors in several applications: an intentional naming scheme developed for the MIT Oxygen project, a constraint solver for first-order relational logic, and a fault-tree analysis system developed for NASA. The talk also gives an overview of the technologies that are being built to enable development of reliable software by the Software Verification Validation and Testing Group at the UT's Center for Excellence in Distributed Global Environments. |
|
