Sign In

Communications of the ACM

ACM TechNews

Deep Learning For Science: A Q&a With Nersc's Prabhat

View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
NERSC's Prabhat

"The broader science community is waking up to the possibilities of machine learning to help them solve . . . problems," says Prabhat, who leads NERSC's Data and Analytics Services Group.

Credit: NERSC

In an interview, the U.S. National Energy Research Scientific Computing Center's (NERSC) Prabhat discusses deep learning, machine learning, and the challenges of applying them to science.

"I think of deep learning . . . as a subset of machine learning, which in turn is closely related to the field of statistics, and all of them have to do with solving inference problems of one kind or another," Prabhat says. He cites the accessibility of big data, more powerful computers, and their convergence as driving deep learning forward, and notes the preponderance of convolutional network and long short-term memory architectures in deep-learning scientific applications.

Prabhat says NERSC has about 70 users employing deep-learning software, and it has identified and deployed several popular deep-learning frameworks on its Cori system. The challenges Prabhat foresees for deep learning include handling complex and diverse datasets, performance and scaling, and addressing "a lack of theory, interpretability, uncertainty quantification, and the need for a formal protocol."

From HPCwire 
View Full Article


Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


No entries found