GPU computing for statisticiansWelcome to the Iowa State University Statistics Department's fall 2013 seminar series on GPU computing for statisticians. Click the "Talks" tab for the schedule of talks and links to the materials for each talk. Posted materials include lecture slides and example code for all talks, plus screencasts of talks that have already been given.
The purpose of these talks is to teach viewers how to create fast statistical software using GPU parallelism. Participants will understand basic principles of parallel computing and learn tools in R, CUDA C/C++, and Python to help them write GPU-accelerated computer code.
GPUs, or Graphics processing units, were mostly used as gaming toys at their inception. But now, scientists and statisticians use GPUs to speed up their data analysis software by orders of magnitude. Become a part of this revolution in massively parallel computing and come learn how to speed up a variety of parallelizable computations, from Markov chain Monte Carlo methods to K-means clustering and beyond.