Doug Hains

DOUG HAINS

Partition Crossover

I am a PhD candidate at Colorado State University in the Artificial Intelligence Program. My research interests include:

Currently, I am working with the SCHED group on combinatorial optimization problems such as the Traveling Salesman Problem and Satisfiability. State-of-the art heuristic solvers for these problems are quite good at solving uniform randomly generated problems. However, many of these same heuristics suffer from a significant decrease in performance when presented with real-world problems such as those arising in industrial applications. My research focuses on understanding the difficulty of these problems and how we can design better algorithms to solve them.

P300 interface

Previously, I have worked with the CSU Brain-Computer Interface Labratory on the design of a P300-speller, a user-interface that allows subjects to interact with a computer using only their thoughts. Many previous P300-spellers relied on the use of expensive hardware, such as the Biosemi EEG. These costs could be prohibitive to many potential users. One goal of my research was to achieve a low-cost alternative without sacrificing performance. The result of our work is a P300 speller that can achieve similar results as $60,000+ units for approximately $6,000.

Wavefront Tiling

I have also worked with members of the CSU High Performance Computing Group on a parallel version of the Smith-Waterman algorithm for CUDA-enabled GPGPUs. Previous parallelizations of this algorithm for GPUs were inefficient due to excessive global memory usage. By adapting the algorithm to a tiled wavefront strategy, we were able to significantly reduce the number of global memory read and writes, thus improving the overall performance.