CSCI 5828 — Spring 2007

Project Idea: I work, almost on a daily basis, with algorithms for understanding dynamical systems.  These algorithms work, by basically partitioning a large dimensional space into subspaces, and then work is done upon that subspace. For instance, one of the major components of these algorithms needs to calculate the nearest neighbors (ranked list of all, actually) of all points.  One could accomplish this task by breaking up the high dimensional space (and thousdands of points) into regions (like a hypersphere) and then doing the calculation within these regions.  This computation is easily seen to be data parallel.