Big Data: Deep Learning on Mars
Shown above is a hyperspectral image acquired over the Nili Fossae region on Mars by CRISM (Compact Reconnaissance Imaging Spectrometer for Mars) onboard NASA’s Mars Reconnaissance Orbiter, which has been orbiting the planet since 2003. CRISM measures both surface and atmosphere and the boxes represent the result of an algorithm that identifies the areas with the largest mineralogical diversity through a process called unmixing. The result is a set of reflectance spectra that represent the pure components in the terrain. This and other large-scale data sets are being used to develop a broad innovative deep learning framework for new scientific discoveries with wide-scale applications. The interdisciplinary research team is led by UMass Amherst computer scientist Sridhar Mahadevan, Mario Parente (Electrical and Computer Engineering), and Darby Dyar (Mount Holyoke College). Learn more.
Image credit: Mario Parente