Colloquium: Data-Driven Methods for Detecting Surface and Subsurface Geological Features
Data-Driven Methods for Detecting Surface and Subsurface Geological Features
Dr. Youzuo Lin, Los Alamos National Laboratory, USA
With data proliferation in all geoscience domains, novel computational techniques such as machine learning and data analytics are emerging as important research areas in seismology, topography, hydrology, and remote sensing, etc. In this talk, I will cover a couple of my recent work in detecting geological features ranging from subsurface to surface applications. The focus of my presentation will be the development and adaptation of various computational methods to those geoscience problems. Specifically, I will cover computational methods including randomization-based low-rank approximation, Krylov-subspace recycling, and supervised dictionary learning. Two different geophysical measurements will be utilized as testing problems: one is the one-dimensional seismic time-series data sets and the other one is two-dimensional remote sensing imagery data sets. Through those examples, I will demonstrate the great potential that machine learning can bring to the geoscience problems.