ALBUQUERQUE, N.M. — Computers inspired by the human brain are demonstrating an unexpected ability to solve complex mathematical problems, potentially paving the way for a new generation of energy-efficient computing for critical applications like national security, according to research published this week in Nature Machine Intelligence.
Researchers at Sandia National Laboratories, Brad Theilman and Brad Aimone, detailed a novel algorithm that enables neuromorphic hardware to tackle partial differential equations (PDEs). PDEs are the mathematical foundation for modeling phenomena ranging from fluid dynamics and electromagnetic fields to structural mechanics, and traditionally require immense computational power to solve.
“You can solve real physics problems with brain-like computation,” Aimone said. “That’s something you wouldn’t expect since people’s intuition goes the opposite way. And in fact, that intuition is often wrong.”
The advance comes as the Department of Energy and the National Nuclear Security Administration (NNSA) seek more efficient methods for running physics simulations. Supercomputers currently used within the nuclear weapons complex consume substantial amounts of electricity. Neuromorphic computing offers a potential path to significantly reduce energy consumption although maintaining computational performance.
For years, neuromorphic systems – which process information in a manner analogous to the brain – were primarily considered suitable for tasks like pattern recognition and accelerating artificial neural networks. The ability to handle mathematically rigorous problems like PDEs was not widely anticipated.
Theilman, a postdoctoral appointee at Sandia whose research focuses on applying neuroscientific principles to neuromorphic computing, explained that the team’s success wasn’t entirely surprising. “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,” he said.
Theilman earned his Ph.D. In computational neuroscience from UC San Diego in 2021, focusing on topological approaches to understanding neural population activity in songbird brains. He and Aimone based their circuit design on a well-established model within the computational neuroscience field, revealing a previously unrecognized link between the model and PDEs.
“We based our circuit on a relatively well-known model in the computational neuroscience world,” Theilman said. “We’ve shown the model has a natural but non-obvious link to PDEs, and that link hasn’t been made until now — 12 years after the model was introduced.”
The research as well has implications for understanding the brain itself. Aimone suggested that the brain routinely performs complex calculations without conscious awareness. “Pick any sort of motor control task — like hitting a tennis ball or swinging a bat at a baseball,” he said. “These are very sophisticated computations. They are exascale-level problems that our brains are capable of doing very cheaply.”
The researchers believe their work could contribute to a better understanding of how the brain processes information, potentially offering insights into neurological disorders. “Diseases of the brain could be diseases of computation,” Aimone said. “But we don’t have a solid grasp on how the brain performs computations yet.”
The Sandia team is now exploring whether their algorithm can be extended to solve even more advanced mathematical problems. The research was funded by the Department of Energy’s Office of Science and the NNSA.
