Learning to Learn: Designing Novel Neuromorphic Algorithms with Machine Learning

Scientific research generates data faster than can be analyzed with today’s computing power, creating a need for energy efficient, intelligent computing.  An emerging technology called neuromorphic computing shows promise in satisfying this need.  For the U. S. Department of Energy (DOE), this breakthrough computing paradigm will be necessary in future computing systems to continue scientific discoveries and development.  A significant research challenge in applying neuromorphic computing to scientific research is defining the learning approaches that will fully realize the real-time continuous learning capabilities of neuromorphic computers.  To address this challenge, machine learning with existing high-performance computing (HPC) will be used to automatically create new learning algorithms that will enable real-time continuous learning for neuromorphic systems.