Machine learning provides powerful tools for predictive analytics - given historical data
predict future behavior of a system. But which ML model to pick? Deep Learning? Boosted
Trees? Support Vector machines? Old fashioned regression? There are too many possible
techniques for casual exploration. The talk will present a systematic approach to model
selection and hyper-parameter search on HPC. A useful toy problem developed for multiple
sensor inference will be presented as a motivating problem and we will walk through Jupyter
Notebooks that the audience can download and use for their own research.

Biography:

Dr. Andrew Strelzoff graduated with a PhD in Computer Science from the University of
California Santa Barbara in 2004. Dr. Strelzoff was an assistant and then associate professor
and head of the computer science program at USM and then joined ERDC Information
Technology Lab in Vicksburg in 2013. At ERDC as part of the Data Science team and lead on
several projects Andrew works on Machine Learning on HPC.