Computer Science Colloquia
Overview:
Thursday, September 6th, 2007,
4:00pm
Windstream Room, Hardymon Building
The Role of Trust and Criticality in Shared Computing Environments
J. Bryan Lyles
Telcordia
Abstract:
Configuration of network infrastructure has traditionally been done manually. This does not scale as systems become more diverse. In shared utility computing environments such as cluster farms or blade-servers there may be programs from different organizations that have been subjected to differing levels of scrutiny. This talk will describe how reasoning about an application's criticality and how much it is trusted can be used to configure intra-cluster protection mechanisms. We will then discuss how these concepts can be used to choose appropriate VPN configurations based on risk assessments and application criticality.
Thursday, September 13th, 2007,
4:00pm
Hardymon Building
Ray-space Caustics and Curvatures
Jingyi Yu
Unversity of Delaware
Abstract:
Caustics are important visual phenomena, as well as challenging global illumination effects in computer graphics. Physically caustics can be interpreted from one of two perspectives: in terms of photons gathered on scene geometry, or in terms of a pair of caustic surfaces. These caustic surfaces are swept by the foci of light rays. In this paper, we develop a novel algorithm to approximate caustic surfaces of sampled rays. Our approach locally parameterizes rays by their intersections with a pair of parallel planes. We show neighboring ray triplets are constrained to pass simultaneously through two slits, which rule the caustic surfaces. We derive a ray characteristic equation to compute the two slits, and hence, the caustic surfaces. Using the characteristic equation, we develop a GPU-based algorithm to render the caustics. Our approach produces sharp and clear caustics using much fewer ray samples than the photon mapping method and it also maintains high spatial and temporal coherency. Finally, we present a normal-ray surface representation that locally parameterizes the normals about a surface point as rays. Computing the normal ray caustic surfaces leads to a novel real-time discrete shape operator.
Monday, September 24th, 2007
Faculty Colloquia: Brief Introductions to Some of Our Research
Two Approaches to Morphology
Raphael Finkel
Abstract:
I will describe and demonstrate with programs two strategies for automatically generating the morphology of Latin verbs. The first is to hand-code default inheritance hierarchies in the KATR formalism. The high degree of similarity among verbs of different conjugation classes allows us to formulate general rules; these general rules are, however, sometimes overridden by conjugation-specific rules. The second is to start with a paradigm chart, then abstract out common parts, then use computational tools to remove redundant morphosyntactic columns, deduce principal parts, combine similar conjugations, and then generate KATR formalism that produces a chart for a set of lexemes. As a side effect of this analysis, we discover that Latin verbs can be described by four principal parts, in consonance with years of pedagogic experience.
The Conference Paper Assignment Problem
Judy Goldsmith
Abstract:
Most artificial intelligence conferences choose their program committees with the expectation that the program committee members will review all submitted papers. Each paper gets several reviewers, who rate and—if necessary —discuss the papers. The question this work addresses is, how to best assign papers to reviewers. In some cases, reviewers mark their expertise on a fixed set of key words, before seeing papers. In some cases, reviewers "bid" on papers, indicated their level of desire to review those papers. The question is then, how to keep the reviewers from complaining too much about their assignment.
I will survey a variety of problem formulations and solutions, and leave
you convinced that the problem is interesting.
This is joint work with Robert Sloan, University of Illinois-Chicago,
and may also mention work by Patrice Perny, Jean-Mathieu Segura, nd
Olivier Spanjaard, Laboratoire Informatique Paris VI, Universite
Pierre et Marie Curie - Paris VI.
Thursday, September 27th, 2007,
4:00PM
Windstream Room, Hardymon Building [info]
Thacher lecture
WHAT DO THOSE IMAGES HAVE IN COMMON?
Unsupervised Learning of Categories Appearing in Images
Narendra Ahuja
University of Illinois at Urbana-Champaign
Abstract:
This talk is about discovering and modeling previously unspecified object categories that happen to appear in a given set of arbitrary images. It is thus about discovering and articulating the dominant themes in the images. Specifically, given a set of images containing frequent occurrences of objects from multiple categories, the objective is to learn a compact model that captures the intrinsic, image-space nature of the categories as well as their relationships, for the purposes of later recognizing and extracting any occurrences in new images. The categories are not defined by the user, indeed they are not even known to the user a priori. Whether and where instances of any categories appear in a specific image is also not known.
The methodology we have developed can also be used for detecting recurring
image themes of more general kind. An example is that of identifying and
extracting the stochastically repeating parts of visual texture, commonly
called as texture elements (e.g., waterlilies on the water surface, a
housing development, crowd of people).
Monday, November 12th,
2007
Joint CS & Cognitive Science Talk
Scientific Data Mining: The Discovery and Use of Complex Networks in Neuroscience and Genomics
Terran Lane
University of New Mexico
Abstract:
Modern science is overwhelmed by a sea of data. Recent years have
brought us sensor technologies that produce gigabytes to terabytes of
information per experiment: functional neuroimaging technologies,
genetic microarrays and high-throughput assays, digital telescopes,
and environmental sensor networks, to name just a few. These
technologies offer unprecedented opportunity for scientific discovery
to the domain scientists. Yet at the same time, they present a
daunting analysis task: to extract meaningful, substantiable patterns
from this overwhelming mass of data. Further, the data are typically
extremely noisy and the patterns of interest are often multivariate
and nonlinear.
To address these analysis problems, computer scientists in the machine
learning and data mining communities have been developing the field of
scientific data mining: using advanced computational and
statistical tools to extract complex patterns from large, difficult,
scientific data sets.
In this talk, I will give an overview of my recent work on scientific
data mining in two different domains: neuroscience and genomics. On
the former front, I will discuss the problem of network
identification: finding the network of functional activity
interactions that underlies some behavioral pattern. The ability to
find such networks is critical to neuroscientists who are working to
understand mental illnesses such as dementia or schizophrenia. On the
latter front, I will discuss the task of biological parameter
estimation for RNA interference (RNAi). In this case, we use the
structure of known activity networks to infer parameters of the
biological process that produced it. These parameters, in turn, help
biologists and pharmacists develop better RNAi-based genetic screens
and pharmaceuticals.
Friday, January 11th, 2008
(Cosponsored by the Cognitive Sciences Program)
Understanding Human Behavior from Low-Level Sensor Data (323 RMS)
Henry Kautz
University of Rochester
Abstract:
The convergence of advances in algorithms for probabilistic reasoning and the development of low-cost, easily-deployed sensors is reviving the dream of AI to develop systems that can understand the narrative of ordinary human life. On the reasoning side, the AI community is developing techniques that bridge the gap between propositional Bayesian representations and hierarchical models of goals, plans, and actions. On the sensing side, new technologies such as RFID tags, GPS, motes, and wearable multi-modal sensors allow us to gather direct information about various aspects of human experience. This talk will describe research on methods for learning probabilistic models of human performance of everyday high-level tasks, and applications to assistive technology and automated life diaries.
Bio:
Henry Kautz is a Professor in the Department of Computer Science at the University of Rochester and was founding director of the Intelligent Systems Research Center at Kodak Laboratories. He came to Rochester in the fall of 2006 after a career at the University of Washington, Seattle, and AT&T Bell Laboratories. He was recently elected a Fellow of the American Association for the Advancement of Science.
February 7, 2008 at 4:00 pm,
CB219.
Semantic web policies for security and privacy
Piero Bonatti
IRIT
Abstract:
Today's most "interesting" systems are open to the internet, so they are vulnerable to a number of possible attacks against system security and their users' privacy. Traditional countermeasures generally conflict with usability.
Trust negotiation (TN) comprises a family of new, appealing techniques for achieving a better tradeoff between (i) security and privacy protection, and (ii) simple and friendly data and service publication on computer networks. According to the current approaches, the peers involved in a transaction acquire a suitable level of trust in each other by negotiating and exchanging electronic credentials and other (digitally unsigned) information. This technique is very flexible, but it requires solving nontrivial problems.
Some of them concern usability and interoperability. Automated support to credential exchange is necessary to preserve smooth navigation experience, but this requires heterogeneous peers to understand each other's requirements. Here is where lightweight semantic techniques come into play.
Moreover, it can be argued that no sophisticated framework for protecting security and/or privacy can reach its full potential unless end users can understand and control it. So a major open problem consists in designing effective automated explanation systems and friendly policy specification languages. The specificities of TN frameworks permit a specialized approach that may result more effective than generic explanation approaches.
In this talk we shall give a brief overview of the above research topics as they are being tackled - with a Semantic Web perspective - within the European network of excellence REWERSE (REasoning on the WEb with Rules and SEmantics) and implemented in the trust negotiation framework Protune.
February 20, 2008, 4:00 pm,
Windstream Room,
Hardymon Blg.
Cryptography (Security Series)
A. Klapper
University of Kentucky
Abstract:
Software systems that provide security use an assortment of basic security protocols such as secure message passing, authentication, and access control. Cryptology is the mathematical study of such protocols. It is based primarily on algorithm design, complexity theory, number theory, and combinatorics. In this talk I will survey the basic tools: private key encryption -- block ciphers and stream ciphers; public key encryption -- e.g., RSA and discrete log systems; and digital signatures. I will also describe various crucial low level primitives such as key generation, hash functions, message authentication codes, highly nonlinear Boolean functions, zero knowledge proofs, and secret sharing schemes. Cryptology has a Dark Side -- cryptanalysis, the craft of attacking cryptosystems. I will talk a bit about some general methods of attack and how they have driven cryptology.
January 23, 2008, 4:00pm, Windstream
Room, Hardymon Blg.
"Please stare at this spot and hold your head very
still", On Biometrics and Security
K. Calvert
University of Kentucky
Abstract:
Traditionally, authentication happens via "something you have, something you know, something you are". This talk will be an overview the last of these three. After stating the problems that biometrics are proposed to solve, I will describe generic system architectures. A description of several popular (and emerging) biometric mechanisms will be followed by a discussion of the performance of various methods, including any apparent impediments to practical deployment.
March 26, 2008, 4:00pm, Windstream
Room, Hardymon Blg.
Network Security: Is it needed and who can you trust?
James Griffioen
University of Kentucky
Abstract:
Some have argued that end systems, not the network, need to be secured. This talk will discuss what it means to secure "the network", whether it is needed, what the challenges are, why trust is important/hard, which protection mechanisms exist today, and how the evolving definition of "the network" is creating new issues to address.
March 28, 2008, 3pm, 323 RBM
(formerly known as Robotics Blg.)
Sequential Voting in Multi-Issue Domains
(Cosponsored by the Cognitive Sciences Program)
Jerome Lang
IRIT
Abstract:
In many real-world group decision making problems, the set of alternatives is a Cartesian product of finite value domains for each of a given set of variables (or issues). Dealing with such domains leads to the following well-known dilemma: either ask the voters to vote separately on each issue, which may lead to so-called multiple election paradoxes as soon as voters' preferences are not separable; or allow voters to express their full preferences on the set of all combinations of values, which is practically impossible as soon as the number of issues and/or the size of the domains are more than a few units. We try to reconcile both views and find a middle way, by relaxing the extremely demanding separability restriction into this much more reasonable one: there exists a linear order x_1 less than ... less than x_p on the set of issues such that for each voter, every issue x_i is preferentially independent of x_{i+1},..., x_p given x_1, ... x_{i-1}. This leads us to define a family of sequential voting rules, defined as the sequential composition of local voting rules. These rules relate to the setting of conditional preference networks (CP-nets) recently developed in the Artificial Intelligence literature. Lastly, we study in detail how these sequential rules inherit, or do not inherit, the properties of their local components.
April 1, 2008, 3:00pm, Windstream Room,
Hardymon Blg.
Scalable Centralization of Content Distribution via Application-Agnostic
Bandwidth Conservation
Aaron Striegel
University of Notre Dame
Abstract:
The centralized approach to content distribution has long been the exclusive purview of resource-rich content providers (Google, eBay, etc.) or the mark of unpopular content. Despite its allure with regards to content customization or safeguarding sensitive user data, content providers typically employ content distribution networks (CDNs) to meet their scaling and performance needs. The mutual exclusion of centralization and scalability is nearly unquestioned dogma, i.e. distributed schemes are nearly always better. In this talk, I will discuss our work that shows that centralization and scalability can indeed play nice together without application- specific proxying. Our work, ScaleBox, operates transparent to the client and network core offering scalable, centralized content distribution in the current Internet. To provide its performance, ScaleBox blends the novel contributions of stealth multicast, enhanced packet caching, TCP acceleration, and tail synchronization in an application-agnostic manner. The talk will give a detailed overview of ScaleBox and its components through both theoretical and general discussions. In particular, I will focus on the packet caching / TCP acceleration components along with a new approach for content-driven data addressing for improving broadband connectivity. As time allows, I will briefly discuss our current implementation efforts regarding ScaleBox in both UNIX and the Intel IXP network processor.
April 3, 2008, 3:00pm, Windstream Room,
Hardymon Blg.
Computer Vision under Programmable Lighting: from 3D Reconstruction to
Object Recognition
Li Zhang
University of Wisconsin
Abstract:
Traditional computer vision analyzes images taken under homogenous lighting conditions, be they indoor or outdoor. With the advent of digital projection technology, we can easily illuminate an environment with programmed spatiotemporal lighting patterns. Such a lighting environment provides new opportunities for computer vision research. I will overview three recent projects that analyze images taken under this type of lighting environment. First, I will present a temporal defocus method, which reliably recovers the 3D structure of a scene, regardless of its occlusion complexity. Then, I will present a spacetime stereo method which accurately reconstructs objects that are deforming over time. Both methods significantly outperform the state-of-the-art techniques for 3D sensing. I will demonstrate several applications of the proposed methods to computer graphics, including image refocusing, video composition, expression synthesis, and facial animation. Finally, I present a method of using projected light to create optical ("virtual") tags in a scene. Each tag carries information about its own 3D location and the identity of the object it falls on. Such a tagging method enables us to detect and recognize not only objects that visible in an image but also the ones that are occluded or slightly outside the field-of-view of the image.
April 30, 2008, 4:00pm, Windstream Room,
Hardymon Blg.
A recent snapshot of network security: problems, policies,
solutions, and unanswered questions
Saikat Chakrabarti
University of Kentucky
Abstract:
Digital information in the business world today is inescapably connected. People depend on computers and communication, ranging from networks (the Internet) for electronic mail, to systems that monitor the nation's critical infrastructure. However, the news is replete with stories of vulnerabilities in network systems that had been exploited by malicious software. Tangible security that can be implemented in the "real world" transcends the mathematical formulations of cryptography. The process of security involves human users and networked-machines and the interactions between them. Security protocols need to be efficient, fault tolerant, and scalable before they can be put to commercial use. In this talk, we will discuss some examples of recent security breaches and network hacks. We will survey important issues in network security, and present what we--Network Security folks at CS.UK--are doing to address these security problems. We will end the talk with questions that remain unanswered regards designing functional network security protocols.
May 15, 2008, 2:00pm,
Center for Visualisation and Virtual Environments,
1 Quality Street, Suite 800, Lexington , KY 40507
Title: ShapePalettes: a novel approach for 3D markup
Michael S. Brown
National University of Singapore
Abstract:
This talk overviews a simple interactive approach to specify 3D shape in a single view using "shape palettes". The interaction is as follows: draw a 2D primitive in the 2D view and then specify its 3D orientation by drawing a corresponding primitive on a shape palette. The shape palette is presented as an image of some familiar shape whose local 3D orientation is readily understood and can be easily marked over. The 3D orientation from the shape palette is transferred to the 2D primitive based on the markup -- only sparse markup is needed to generate expressive and detailed 3D surfaces. This markup approach can be used to model freehand 3D surfaces drawn in a single view, or combined with image-snapping tools to quickly extract surfaces from images and photographs.
Bio:
Michael S. Brown received his BS (1995) and PhD (2001) from the University of Kentucky and was visiting PhD student at the University of North Carolina from 1998-2000. He has held previous assistant professor positions at the Hong Kong University of Science and Technology, California State University - Monterey Bay, and Nanyang Technological University. He is currently the Sung Kah Kay Assistant Professor in the School of Computing at the National University of Singapore. His research interesting include Computer Vision, Image Processing, and Computer Graphics.