Optimizing on Predictions from Machine Learning Pipelines
When: Monday, September 23, 2019
Where: PGH 563
Time: 11:00 AM
Speaker: Dr. Doug Hakkarinen, ConocoPhillips Analytics Innovation Center of Excellence (AICOE)
Host: Dr. Edgar Gabriel
Supervised learning often focuses heavily on the creation, selection, and tuning of predictive models to meet an acceptable error metric. However, it is typically non-trivial to utilize such a predictive model to drive business value without careful consideration of the business problem being addressed during the construction of the modeling pipeline. This talk will focus on lessons learned during a real-world machine learning project where the objective was to perform constrained optimization on the output of a collection of independent machine learning modeling pipelines.
Dr. Doug Hakkarinen is presently a data scientist in the ConocoPhillips Analytics Innovation Center of Excellence (AICOE). His advanced analytics projects at ConocoPhillips have covered business problems in the US, Canada, Norway, and Australia, spanning machine learning problem domain areas including time series analysis, 3D/4D convolutional neural networks, recurrent neural networks, reinforcement learning, survival analysis, and distributed computing toolsets for data science. Before joining the AICOE, Doug developed high performance computing (HPC) applications for the ConocoPhillips seismic processing group, where he focused on software development for new seismic processing algorithm tools and on improving the performance of distributed seismic processing algorithms.
Doug's work history also includes software development or related positions for Bentek Energy, Tetra Pak, and Sun Microsystems.
Doug has a PhD and MS in mathematical and computer sciences from the Colorado School of Mines, where he was part of the SmartGeo NSF IGERT program. His dissertation work focused on distributed computing and optimization algorithms to support geophysical applications. Doug also holds Bachelor's degrees from the University of Colorado at Boulder in Molecular Biology and Computer Science.