Mount Holyoke College professor Darby Dyar has teamed up with researchers from the University of Massachusetts Amherst to apply biologically-inspired, “deep-learning” methods to computers to analyze large amounts of scientific data from Mars. By doing so, the researchers will learn more about the surface chemistry of the planet, according to an article in Sciencemag.
The National Aeronautics and Space Administration’s Mars rover, Curiosity, uses laser-plus-spectrometer instruments to analyze rocks—the same technique Dyar, chair of the astronomy department, employs in her lab at Mount Holyoke. Curiosity has been sending a steady stream of specialized camera images back to earth from its Martian crater since August 2012. The data range from one-dimensional spectra of rock samples to three-dimensional hyperspectral images of the Martian surface.
Earth-side analysis involves employing computers to learn from the data they analyze and to recognize patterns as they emerge—a technique known as deep learning.
“The combination of deep-learning methods and burgeoning data production made possible by modern instrumentation is going to revolutionize nearly all applications of spectroscopy,” said Dyar.
“This summer, five undergraduate students are working to finish collecting the largest known data set of laser-induced breakdown spectra of geological materials in the world, with nearly two billion data points,” she said. “Deep-learning methods and modern machine-learning techniques are absolutely critical to interpreting such a large mass of data.”
Read the whole story.