SCU Mathematics Colloquium Series Schedule
Title: Acquiring Computer Graphics Models of People
Abstract: Computer generated imagery and 3D models are a core component of modern scientific analysis, entertainment, and communication of ideas. Unfortunately, most methods for creating sufficiently complex models rely heavily on tedious manual specification and skilled artists. Acquiring models directly from the real world allows us to leverage and explore the phenomenal complexity of nature. this talk will present the results of several projects for recovering computer graphics models directly from the real world. Space-time stereo generalizes several recovery methods into a single framework. This has allowed the shape of deforming objects to be recovered at a higher quality than was previously possible. I'll also discuss constructing an animatable model of the space of all human shapes and poses. we scanned a bunch of people at high resolution, brought the models into correspondence, and then built a model which predicts body surface shape as a function of identity and pose.
Title: The Actuarial Profession - How to Prepare Yourself Now
Abstract: The actuarial profession is becoming a more and more popular career choice for recent graduates - so the competition for jobs is becoming more fierce. This means that students must now do everything they can to prepare themselves (and their resume) for the job search ahead. This talk will focus on the Casualty Actuarial Society (CAS) examination process, what employers look for in entry-level candidates, and the type of people that might find this profession rewarding as a long-term career choice.
Title: Extended UPGMA for Phylogenetic Tree Reconstruction
Abstract: We study the UPGMA algorithm for equidistant phylogenetic tree reconstruction using the geometry of the space of all equidistant trees with n leaves, also known as the Bergman complex of the graphical matroid for the complete graph K_n. We show that this classic algorithm performs an orthogonal projection of the data (empirical distance data among n taxa) onto a distinguished cell of the Bergman complex. We also show that the equidistant tree with the least (Euclidean) distance from the data is obtained from such an orthogonal projection, but not necessarily given by the UPGMA algorithm. We give two extensions of this algorithm which use the geometry of the Bergman complex and demonstrate the computational results from our implementation. The first one is an algorithm that rapidly produces a tree (usually better than the UPGMA tree) for as many as 100 taxa. The second one is an exact algorithm that finds the best least square tree for up to 20 taxa in reasonable time.
Title: Ecological Models and their Implications