February 16 - A Gentle Introduction to Statistical Thinking
Associate Professor of Mathematics
Abstract: H.G. Wells said "Statistical Thinking will one day be as necessary for efficient citizenship as the ability to read and write." Computers, which have allowed us to collect and store information in unimaginable quantities, have thrust us into the Information Age. Statistics provides a method of wading through this data to gain insight and draw conclusions. In this talk, geared towards those with no statistical knowledge, we discuss the basic principles of statistical reasoning.
February 23 - Breaking up the Internet Transport Logjam
Assistant Professor of Mathematics
Franklin & Marshall College
Abstract: The Internet's traditional transport services architecture worked well in the limited heterogeneity of the early Internet, but fails to statisfy the more diverse demands of today's users, administrators, applications, and physical networking technologies. While the need for new transport services continues to persist, new transport protocols are almost impossible to deploy in the current Internet. I will identify the architectural causes of this evolutionary logjam, and I will discuss our proposal for Tng ("Transport next generation"), a new transport services architecture that modularized orthogonal transport functions, allowing each function to be enhanced or specialized without interfering with other functions in the Internet's Transport Layer.
No background in networking is expected for this talk --- I will discuss the necessary background --- but a healthy sense of curiosity is required. The ability to throw peanuts from the gallery is a plus.
March 2 - A Recurring Theme in the Mathematics of Sports
Professor of Mathematics
Abstract: In anticipation of April as Mathematics Awareness Month, this presentation will focus on this year's theme, "Mathematics and Sports." One of the primary themes of a discrete mathematics course is "recursive thinking." As it turns out, it is easy to use this idea in the modeling of many games and sporting events, ranging from baseball to bowling to tennis. In this presentation, we will start with simple recursive models and build up to Markov chain analysis. Along the way, we will compare the predictions of some of our models to real data from the world of sports and games. See http://www.mathaware.org for more information on Mathematics Awareness Month.
March 9 - Towards Open-Domain Semantic Parsing
Abstract: Semantic parsing, or the automated act of converting linguistic utterances into formal representations of their meaning, has been a goal of the Artificial Intelligence community from its beginning. Recently developed learning techniques have produced highly encouraging progress in building accurate syntactic parsing systems, and to a more limited degree, semantic parsing systems. This talk describes work that addresses two shortcomings of the existing supervised-learning paradigm for semantic parsing: first, classifiers today are still tied to the domain of their training data, with significant drop-offs in performance when they are applied to new domains. In response, we have developed novel, unsupervised representation-learning algorithms for porting statistical techniques for semantic parsing to new domains, without requiring any new labeled data from the new domain. Second, statistical techniques for semantic parsing have ignored the fact that world knowledge is necessary for disambiguating many linguistic utterances. He will describe novel techniques for extracting world knowledge from text, and using that knowledge to help understand sentences.
March 23 - "Sketch-based Bargello Quilt Design"
Instructor of Computer Science
Traditional computer-aided design systems excel in high-precision domains, but their expense and steep learning curves make them inaccessible to most lay people. Sketch-based design is a complementary computer-assisted design paradigm that allows users to generate designs quickly and easily by sketching with a mouse or stylus. The challenge is to provide systems that are easy to learn and use, but that nevertheless allow the creation of content-rich designs. To accomplish this, sketch-based systems focus on specific design domains, incorporate knowledge from those domains, and use it to infer complex designs from simple sketches. We present a sketch-based system for designing Bargello quilt patterns, which respects the physical constraints of the Bargello style while supporting users in a design process that is significantly easier than the manual process in common use today.
March 26 - "Scrambled Bits and Hash"
Dr. John Sigle
Louisiana State University in Shreveport
Encryption techniques play a major role in network security. Both traditional (shared-key) and public-key encryption techniques will be discussed with a focus on how they are used to provide secrecy and authentication. Digital signatures provide authentication, message integrity and non-repudiation. The use of hash functions and message digests to implement digital signatures will be described.
April 6 - "The Coolest Theorem in Computer Science: Computers Can't Do Everything"
Assistant Professor of Computer Science
Abstract: You might think it's obvious that computers can't do everything, but could you prove it? Alan Turing did, in 1936, before the first electronic computer had even been built! This talk will explain how he did it. No previous knowledge of computer science or mathematics is necessary.
April 13 - "Medical Imaging and Biostatistics"
Jeff Goldsmith, '07
Doctoral Student in Biostatistics
John Hopkins School of Public Health
Abstract: If there's one thing that statistics thrives on, it's data. Recent technological advances have allowed doctors to acquire and store massive amounts of information with the hope of using that information to improve patient care. In this talk (geared towards an audience with little or no statistical background) we'll discuss recent work in the analysis of MRI and SPECT images, and go on to look at the use of neuroimaging in predicting adverse outcomes for multiple sclerosis patients.
May 4 - Honors Presentation - James Doyle
"PathFinder in CUDA"
Image segmentation into superpixels is a common early step in computer vision algorithms. Several algorithms already exist that produce good quality results. PathFinder is a competing image segmentation algorithm that creats high quality superpixels at faster speeds. Our goal is to transfer computation from the CPU to the GPU, which has the potential to execute parallel tasks much more efficiently. We investigate CUDA, a freely available platform for programming GPUs, and learn what kinds of operations benefit most from this approach. We find a 3 to 5 times speed up of the PathFinder algorithm overall, with an improvement of three orders of magnitude in some components of the algorithm.
May 6 - Honors Presentation - Michael Keating
"Executing Formal Specifications via Constraint Programming: Enhancing the jmle Tool"
The Java Modeling Language (JML) is a language that can be used to embed formal specifications into Java programs. Because formal specifications are intrinsically abstract, they cannot typically be executed. The jmle tool, created by Professor Tim Wahls, enables the execution of JML formal specifications by converting them into executable constraint programs.
The goal of this year-long research project is to improve the functionality of the jmle tool with respect to solving finite domain variables in constraint programs. I have implemented new constraints that remove invalid integers from every involved variable's finite domain before the backtracking process begins, greatly reducing the number of variables in division and array constraints faster and more efficiently than ever before. The success of this research demonstrates the potential to further enhance the jmle tool's performance when solving finite domain constraints.
Location: Tome 117