Note: Contributions from CS/STAT/MATH faculty speakers would be much appreciated.
The CAVE is a 10x10x9 foot room in which stereoscopic glasses are worn to view 3D images on four walls. We will describe the CAVE virtual reality system and show how it has been used to study mathematics, including non-Euclidean geometry, topology in three and four dimensions, knot theory, and sphere eversions. Applications involved were shown in VROOM at SIGGRAPH '94 and at SuperComputing '95.
Software Reuse
Will Tracz
Loral Federal Systems
The seminar will focus on the technical and nontechnical issues surrounding software reuse -- at the code and design level. The speaker will relate his extensive background on the state of the art and the state of the practice of software reuse. In particular, he will describe the tools and processes needed to create domain-specific software architectures.
State-Space Models for Species Compositions
Dean Billheimer
University of Alaska Fairbanks
Organisms integrate the environmental factors to which they are exposed, affecting their productivity, reproduction, and mortality. In turn, these biodynamics affect the presence and number of species at a site. Thus, the relative abundance of species (species composition) indicates the ecological condition of the system under study. Data from biomonitoring studies consist of counts of organisms sampled at selected sites and times. The counts often depend on biotic and abiotic factors (such as pH or productivity), and typically exhibit temporal or spatial structure.
A state-space model, based on Aitchison's (1986) model of multiplicative error for proportions, is developed to analyze biomonitoring data. We assume an unobservable `state' composition for each observation of organism counts. The state compositions may be spatially related, and may also depend on covariates. Discrete observations are accommodated by a conditional Multinomial observation model (given the state). Markov chain Monte Carlo (MCMC) is used to provide inference about the species compositions, covariates and spatial dependence parameters. These methods are illustrated in two applications. The first is a designed experiment studying arthropod community response to ecological disturbance. The second is a biomonitoring study evaluating the ecological condition of the Delaware Bay estuary via benthic invertebrate composition.
Bayesian Model Averaging
David Madigan
University of Washington
Standard statistical practice ignores model uncertainty. Data analysts typically select a model from some class of models and then proceed as if the model generated the data. This approach ignores a component of uncertainty, leading to over-confident inferences. Bayesian model averaging (BMA) provides a mechanism for accounting for this model uncertainty. Researchers have recently proposed several methods for implementing BMA. I will discuss these methods and present a number of applications. In these applications, BMA provides sharply improved out-of-sample predictive performance. Several of the applications use Bayesian graphical models, and the talk will describe these models in detail.
Unit vs. Proper for Generalizations of Interval
Ann Trenk
Rutgers
Interval graphs have been studied extensively both because they have many applications and because many problems which are NP-complete for general graphs, admit polynomial-time solutions when restricted to the class of interval graphs. Formally, a graph G is an interval graph if it can be represented as follows: each vertex of G corresponds to a real interval so that two vertices are adjacent in G if and only if their corresponding intervals intersect.
Unit interval graphs are those which have an interval representation in which every interval has the same length. Likewise, proper interval graphs are those which have an interval representation in which no interval properly contains another. Clearly the class of unit interval graphs is a subset of the class of proper interval graphs. In 1969, Fred Roberts showed that these classes are in fact equal.
We study tolerance and bitolerance graphs which generalize interval graphs by allowing a certain amount of overlap of intervals before an edge is formed between the corresponding vertices. We formalize this notion in the talk and discuss which classes have been characterized, which have efficient recognition algorithms, and then focus on the questions of when the unit and proper subclasses are equal.
So You Want to be a Software Professional?
Douglas Canfield
Sound Technologies, Inc.
Technical competence, while important, is far from the most important attribute of a good software professional. Hiring managers look for other traits to determine if a job candidate will contribute to the team effort today, tomorrow and well into the future. The importance of these traits is explored with examples taken from real life experiences.
On the other side of the coin, selecting an employer is even more complex than hiring an employee. When starting a job search, the options can be overwhelming. Suggestions are offered for possible career paths and ways of focusing to find the "right" job. Examples illustrate what it is like to work in some of the available environments. Once you know what you want, it becomes easy to convince a prospective employer that you are exactly the person needed for the job.
Windows NT Architecture and Networking
Alok Sinha
Microsoft Corporation
This talk will present an overview of Windows NT architecture with some comparison to Unix. The networking architecture of Windows NT, internetworking with NT, and remote access with NT will be discussed. NetOLE, the soon-to-be-released distributed object architecture, will also be introduced.