Distinguished Guest Lecturer: Nina Amenta

Nearest-Neighbors on the GPU

There is a renaissance of interest in parallel algorithms now that we have massively parallel graphics processing units and massive distributed systems in the cloud.   We’re considering how easily parallelizable algorithms translate in practice to the GPU, focusing on the nearest-neighbors problem.  Efficient GPU algorithms rely on efficient primitives, mainly scan and sort.  We consider scan-and-sort algorithms in both low and high dimensions, and also the idea of adding a hash primitive.
 

Professor Nina Amenta is the Chair of the Computer Science Department at the University of California at Davis.  Prior to 2002, she was on the faculty at the University of Texas at Austin.  She got her PhD at UC Berkeley in 1993 and an undergraduate degree in Classics from Yale in 1979. 

As a researcher, Professor Amenta works on algorithms that exploit geometry to solve problems in computer graphics; she is best known for using Voronoi diagrams for surface reconstruction and other algorithms in 3D shape representation.   As a teacher, she is interested in the beginning computer science curriculum, recently including Web programming.  As an administrator, she is interested in engineering for sustainability.