CSE 600 (Ongoing Research Seminar)

Abstract:

Rules provide an expressive means for implementing database behavior: They cope with changes and their ramifications. Rules are commonly used for integrity enforcement, i.e., for repairing database actions in a way that integrity constraints are kept. Yet, Rule Triggering Systems fall short in enforcing effect preservation, i.e., guaranteeing that repairing events do not undo each other, and in particular, do not undo the original triggering event.

We suggest a method for enforcing effect preservation on updates in general rule triggering systems. The method derives transactions from rules, and then splits the work between compile-time and run-time. At compile-time, a data structure is constructed, that analyzes the execution sequences of a transaction and computes minimal conditions for effect preservation. The transaction code is augmented with instructions that navigate along the data structure and test the computed minimal conditions. We show that this method produces minimal effect preserving actions, and under certain conditions, provides meaningful improvement over the quadratic overhead of pure run-time procedures.

In this talk we present examples that justify the need for effect preservation, demonstrate our method, prove its essential properties, and analyze its complexity. We also comment about the current implementation and possible extensions to semi-structured data.

Joint work with Steffen Jurk, Brandenburg University of Technology, Cottbus, Germany.

Speaker Bio:

Mira Balaban holds a Ph.D. in computer science from the Weizmann Institute of Science, Israel, and a music degree from the Rubin Academy of Music in Tel-Aviv. She is now affiliated with the Department of Computer Science in Ben-Gurion University. Her research is in the areas of knowledge representation, conceptual modeling, database semantics, programming languages, and computer music.

Abstract:

Epitome is a condensed version of an input array of real or discrete values which is of much smaller dimensions and yet it "epitomizes" the input data in terms of of its pieces. One way to think of an epitome is as a small mosaic assembled from the data pieces. As the epitome is much smaller than the input data, it is especially well suited to the data containing many repeating patterns with moderate variations. Epitome can be viewed as a generative model of data subarrays and estimated using machine learning techniques, leading to interesting interpretations in various applications. In particular, I will describe epitome as a representation of images suitable to layer extraction, as well as a basis of a human immunodefficiency virus (HIV) vaccine.

Speaker Bio:

Nebojsa Jojic is a researcher in the Machine Learning and Applied Statistics Gorup at Microsoft Research, which he joined in 2000. He received his PhD in electrical and computer engineering from the University of Illinois at Urbana-Champaign in 2001 where he was awarded Robert T. Chien award for excellence in research in 2000 and a Microsoft Graduate Fellowship in 1999. Dr. Jojic has been working on statistical inference problems in various domains including computer vision, audio processing and computational biology.

Abstract:

Spline surfaces have been broadly applied in all engineering fields. The most popular Spline schemes are still defined on planar regions. It is a long lasting open problem to define Spline surfaces directly on manifolds with arbitrary topologies.

Our recent discovery builds a relation between manifold spline and affine atlas of the manifold. John Milnor showed that the only manifolds which admit affine atlas are tori. Therefore, the intrinsic difficulty of defining manifold spline is caused by the topological obstruction, which can be explained using characteristic class theory.

By removing Euler number of extraodinary points, any manifold admits an affine atlas. Such an atlas can be explicitly constructed using holomorphic 1-form from Riemann surface theory.

The algorithm for constructing manifold spline on general surfaces will be explained in the talk. Some examples of modeling shapes using manifold splines will be shown.

Speaker Bio:

Xianfeng David Gu got his PhD in computer science department, harvard university. In his PhD thesis, Dr. Gu solved an open problem for computing Riemann surface structure for general surfaces under the supervision of Shing-Tung Yau, a Fields medalist. He also developed the technique of "geometry image", which unifies geometry and image processing techniques. Currently, his research focuses on manifold spline, computational Riemann surface theory and computational optics.

Abstract:

An overarching concern in software security is the need to work with the large base of existing software. Approaches that "rip out" today's software to replace it with software that is "secure by design" are generally too expensive to be practical. As a result, research in software security has been focussed on compiler and OS level techniques that can transparently add more security to systems, without requiring them to be redesigned or reimplemented. A number of techniques that have been developed in the context of program analysis, transformation, and verification, and operating systems have formed the basis of exciting new developments that promise to make today's commodity software more secure. In this talk, I will provide a brief survey of the research landscape in software security, and follow up with a description of some of the projects that are being carried out in the Secure Systems Laboratory.

Speaker Bio:

R. Sekar is an Associate Professor of Computer Science and the director of the Center for Cybersecurity at Stony Brook University. His recent research emphasizes proactive cybersecurity, where security problems are addressed before they are exploited, and system survivability, i.e., the ability of an information system to carry out its mission-critical tasks even in the face of attacks. His technical interests include language and formal methods based security, operating systems, machine learning and anomaly detection, and usable security. His research has been supported by several grants from NSF, DoD and local industry. Prior to joining Stony Brook, Sekar got his Bachelor's degree in EE from IIT, Madras and a Ph.D. in CS from Stony Brook in 1991. From 1991 to 1996, he was a Research Scientist in Networking Research at Bellcore. He then accepted a faculty position at Iowa State University, and subsequently moved to Stony Brook in 1999.

Abstract:

I will start by giving a brief overview of the field of natural language processing. I will then talk about a piece of research that exemplifies the current tension between knowledge-driven and data-driven approaches to natural language processing. This research was joint with researchers from AT&T and the University of Pennsylvania. I will conclude with a summary of opportunities for students who wish to know more about NLP.

As spoken dialog systems have become more complex, there has been increasing interest in the use of spoken language generation (SLG). SLG promises portability across application domains and dialogue situations through the development of application independent linguistic modules. However, rule based SLGs often have to be tuned to the application to improve efficiency and output quality. In this talk, I will describe the use of machine learning to adapt a sentence planner to generate high quality information presentations in the multimodal dialogue system MATCH.

Speaker Bio:

Dr. Amanda Stent has been at Stony Brook for almost three years. She previously worked briefly at AT&T Research. She got her Ph.D. in 2001 from the University of Rochester. She currently supervises or co-supervises five PhD students, three MS students and three undergraduates. Her research focuses on natural language and multimodal generation and dialog systems.

Abstract:

The talk discusses the research program in Information Systems Engineering, for which the principal domains are medical informatics and marine sciences. In medical informatics an underlying theme is the reengineering of medical processes to improve the quality of health care, especially the reduction of medical errors. This reengineering first addresses the replacement of manually provided clinical information, typically contained in a medial chart, with electronic data. Once the processes are automated, new techniques can be developed for the presentation and analysis of medical data. Two of the techniques developed to date will be discussed: electronic representation of clinical practice guidelines and synchronization and combinations of medical monitor data. The marine sciences projects involve the capture, analysis, and display of marine science data. The projects described are the Long Island Sound system and the shoreline finder system. Both projects capture sensor data (air and water quality for Long Island Sound and images for the shoreline finder system), analyze the data, and present historical trends.

Speaker Bio:

Robert Kelly Associate Chair, Computer Science Department Ph.D. 1991, New York University Rob Kelly's research interests include medical informatics, software engineering, systems engineering, image processing, Internet programming, and parallel programming. The primary research focus has been on systems to reduce the incidents of medical errors in acute health care facilities. This work includes a system for a high level interface to medical monitors, representation techniques for clinical guidelines as system-independent documents, an Anesthesiology clinical support system, a rule-based system for discharge planning, and a system to associate monitor data with outcome data to improve prediction of patient condition in critical care settings. He is also developing systems to process marine science sensor data with the goal of developing inquiry systems for marine features. Rob has published papers on these topics in major journals and conferences. He has been on program committees or chaired invited sessions of a number of international conferences. Rob is also the graduate director of the college level program in Systems Engineering. Collaborating with an industry partner, he has developed a software application that received the Long Island Software Award for Internet-based software.

Abstract:

--

Speaker Bio:

Jessica Hodgins joined the Robotics Institute and Computer Science Department at Carnegie Mellon University as a Associate Professor in fall of 2000. Prior to moving to CMU, she was an an Associate Professor and Assistant Dean in the College of Computing at Georgia Institute of Technology. She received her Ph.D. in Computer Science from Carnegie Mellon University in 1989. She has received a NSF Young Investigator Award, a Packard Fellowship, and a Sloan Fellowship. She was editor-in-chief of ACM Transactions on Graphics from 2000-2002 and Papers Chair for ACM SIGGRAPH 2003.

Abstract:

--

Speaker Bio:

--

Abstract:

--

Speaker Bio:

--

Abstract:

--

Speaker Bio:

--