New research shows that advances in technology could help make future supercomputers far more energy efficient.
Neuromorphic computers are modeled after the structure of the human brain, and researchers are finding that they can tackle difficult mathematical problems at the heart of many scientific and engineering fields.
In a study published in Nature Machine Intelligence, Sandia National Laboratories computational neuroscientists Brad Theilman and Brad Aimone introduce a new algorithm that allows neuromorphic hardware to solve partial differential equations, or PDEs. These equations form the mathematical basis for describing systems such as fluid flow, electromagnetic behavior, and the strength of physical structures.
The results show that neuromorphic systems can not only solve these equations, but can do so with impressive efficiency. According to the researchers, this advance could open the door to the world’s first neuromorphic supercomputer, with major implications for energy-efficient computing in national security and other demanding applications.
A brain-inspired approach to scientific computing
Partial differential equations play a central role in modeling the real world, from forecasting the weather to predicting how materials respond to force. Solving these equations has traditionally required enormous computing power. Neuromorphic computers take a different path, using hardware designs that more closely mirror the way the brain handles information.
“We’re just starting to have computational systems that can exhibit intelligent-like behavior. But they look nothing like the brain, and the amount of resources that they require is ridiculous, frankly,” Theilman said.
