Upcoming seminar speaker sign-up
|To schedule a meeting: Click the 'edit' tab above. Enter your name to the right of the first pair of vertical lines (||) for the desired meeting time and the meeting location after the second pair of vertical lines.
For seminar hosts: This page should be updated in order to reflect the schedule of the current week's seminar speaker. For an empty template of the source text, click here. Customize the schedule to match the speaker's availability, and add events like meals and airport arrival/departure times as necessary.
Click here for the EEB Department Seminar page.
Seminar Speaker: Marc Suchard
Institution: David Geffen School of Medicine, University of California at Los Angeles (UCLA)
Wednesday Seminar Title: "Phylogenetic likelihoods 100-fold faster, or ridiculously parallel statistical inference" 4:00 PM, CLAS 344
Faculty or Student Contact: Paul Lewis
|Important note: Because this is a joint EEB/Statistics seminar, the days of Marc's visit and the location of his seminar are unusual. He will be in the EEB department on Tuesday, but will give his seminar on Wednesday in the CLAS building (home of the Statistics Department).|
Massive numerical integration plagues the statistical inference of partially observed stochastic processes and high dimensional data modeling. An important biological example entertains partially observed continuous-time Markov chains (CTMCs) to model molecular sequence evolution. Joint inference of phylogenetic trees and codon-based substitution models of sequence evolution remains computationally impractical. Parallelizing data likelihood calculations is an obvious strategy; however, across a cluster-computer, this scales with the total number of processing cores, incurring considerable cost to achieve reasonable run-time.
To solve this problem, I describe many-core computing algorithms that harness inexpensive graphics processing units (GPUs) for calculation of the likelihood under CTMC models of evolution. High-end GPUs containing hundreds of cores and are low-cost. These novel algorithms are particularly efficient for large state-spaces, including codon models, and large data sets, such as full genome alignments where we demonstrate up to 150-fold speed-up. I conclude with a discussion of the future of many-core computing in statistics and touch upon recent experiences with massively large and high-dimensional mixture models.
Tuesday, September 21, 2010
|12:00pm||Lunch, with Grad Students||Torrey Life Science (TLS) Bamford Room 171b|
|4:00pm|| Systematics Seminar
|Bamford Room (TLS 171b)|
|6-8||Dinner, Willimantic Brewing Co.||Paul Lewis, Louise Lewis, others?|
Wednesday, September 22, 2010
|8am-4pm||Statistics Department meetings|