January 4, 2017 – From machine learning to neuromorphic architectures that enable greater computing flexibility and utility, Oak Ridge National Laboratory researchers are pushing boundaries with Titan. “We’re using deep learning to advance the state of the art in several challenging fields such as computer vision and speech recognition,” said Steven Young, a member of ORNL’s Intelligent Computing Research Team. Young noted that their approach is providing a promising tool in areas previously unexplored by computer scientists. For example, optimal networks for commercial datasets are vastly different than the optimal networks for scientific data. But by utilizing the 18,688 GPUs on Titan and an evolutionary algorithm, researchers can quickly find the best network for their problem. As a result, scientists are seeing and making sense of data that was previously either obscured or simply unavailable for analysis.
Oak Ridge National Laboratory’s supercomputer is opening new horizons for the Nature Inspired Machine Learning Team.
