Connor Smith

Iterative Solver Parameter Selection: A Machine Learning Solution          

Dr. Mike Heroux, Advisor
Imad Rahal        
Jeremy Iverson
Noreen Herzfeld            

What inspired Connor to select his thesis topic?
My interest in this topic came about from a need my advisor - Dr. Heroux - and I saw in the scientific computing landscape. Tuning the proverbial knobs of large scientific software is a difficult and time consuming endeavor, and I wanted to take much of the guess work out of the equation by introducing automation into the process.    

Connor’s research in his own words:
The goal of my research is to assist professionals in the field of scientific computing as they pick parameters required to run their large pieces of software. I approached this goal from a machine learning perspective, which essentially involves teaching computers to think for themselves. Through machine learning, I was able to prove that computers can do this job better than many of the users of the scientific software, thus demonstrating the immense potential that machine learning represents.