Brad Theilman and Felix Wang, researchers at Sandia National Laboratories, are pushing the envelope in the exciting new field of neuromorphic computing. Their latest work has taken the form of unpacking the neuromorphic computing core to better understand its full potential applications. This cutting-edge technology attempts to replicate some aspects of how the human brain works. Fifth, the brain is incredibly energy-efficient, working on only 10 watts of power.
Because neuromorphic computing excels at processing continuous real-time sensor data it becomes a promising edge computing solution for other applications. The Sandia team is an example of the growing interest developing in this field. Together, they are addressing deep mathematical challenges that have the power to redefine the future of scientific computing.
Advancements in Neuromorphic Hardware
The Sandia team made a splash with their 2021 implementation of the Monte Carlo method on neuromorphic hardware. This approach, key to understanding the theory of convolutions and differential equations, resulted in breakthroughs in computational efficiency. Most recently, they adapted FEM to a model of the motor cortex. They even pulled it off at scale on Intel’s Loihi 2, an ultra-state-of-the-art neuromorphic hardware platform.
The Finite Element Method is perhaps the single largest scientific computing success story. It is indispensable to innumerable engineering and physical science applications, enabling engineers and scientists alike to tackle the most complex problems. By harnessing neuromorphic systems such as Loihi 2, researchers hope to make FEM computations more efficient.
Bradley Theilman commented on the complexity of these tasks, stating, “It’s a complex problem. The brain is controlling muscles in response to real-time information to make contact with the ball.” This piece of wisdom sums up the spirit of the struggle to reproduce human processing in machines.
Potential Energy Advantages
Using neuromorphic hardware for FEM provided substantial energy efficiency benefits. This benefit is particularly pronounced against legacy systems. Neuromorphic computing can lower the power consumption by several orders of magnitude, even while performing these complex mathematical conundrums. From their perspective, scientists hope that neuromorphic systems would provide an energy efficiency edge. That’s largely because of the incredible efficiency at which the human brain solves comparable problems.
James B. (Brad) Aimone, a neuroscientist involved in the research, emphasized the importance of exploring various mathematical problems through this lens. He remarked, “There’s no reason to assume you can’t do something in neuromorphic computing.” His view reflects the larger promise that this technology has—a promise that far too many computer scientists have not fully realized.
Aimone elaborated on the workings of neurons, stating, “Neurons receive weighted information and send a timed, all-or-nothing pulse that is transmitted to near neighbors.” This unique principle guides the functionality of neuromorphic systems, setting them apart from traditional computing approaches.
Collaborations and Future Directions
Steve Furber, a computer scientist emeritus at the University of Manchester, has been one of the field’s most influential figures. He is perhaps most famous for creating SpiNNaker hardware. This platform is ideal for modeling diffusion of a reactive heat source. Most importantly, it shows the potential of neuromorphic systems to be diverse scientific problem solvers.
The partnership among academia, joint institutes and industry underscores the combined resources and creativity required to advance the frontiers of neuromorphic computing. Other teams are realizing the benefits of this technology as well. Consequently, optimism is growing that it will lead to transformative breakthroughs in both theoretical and applied mathematics.