- Global computer science leaders had until yet not taken into account the possibility that brain-like computer systems would perform better on complex mathematical tasks.
- The neuromorphic chips can simulate the discrete-time Markov chains using random walk computations.
In the past decade, researchers across the globe have been putting efforts to design brain-inspired computer systems called neuromorphic computing tools. Today, most systems are being used to execute deep learning algorithms and other Artificial Intelligence (AI) tools.
A research team at Sandia National Laboratories had conducted a study to assess the potential of neuromorphic architectures to execute different types of computations, namely random walk computations. These computations involve a progression of random steps in the mathematical space. The researchers published their findings in Nature Electronics. They recommended that neuromorphic architectures may be best suited for implementing these computations and can, thus, go beyond Machine Learning applications.
“Most past studies related to neuromorphic computing focused on cognitive applications, such as deep learning. While we are also excited about that direction, we wanted to ask a different and complementary question: Can neuromorphic computing excel at complex math tasks that our brains cannot really tackle?” One of the researchers, James Bradley Aimone, said in an interview.
Until yet, global computer science leaders did not consider the possibility that brain-like computer systems would perform better on complex mathematical tasks. The latest study by Aimone and his fellow researchers shows that as against their expectations, this might be the case. Particularly, the team found that the chips might be promising to simulate discrete-time Markov chains using random walk computations.
“We recognized that the brain (and thus neuromorphic computing) has a different type of parallel computing architecture than conventional computers. When we looked across the many types of scientific computing problems, we recognized that Monte Carlo random walks are a particular class of problems that could naturally fit neuromorphic architectures if we were clever about reframing the random walk math to fit these platforms,” Aimone explained.
Members of the research team included mathematicians, computer engineers, and Aimone, a theoretical neuroscientist. With the help of the expertise of each of them, the researchers could examine Monte Carlo simulations that have so far been implemented using traditional computing tools in the context of neuromorphic computing. Due to this, they could bring forth the potential of neuromorphic architectures for accomplishing a well-known complex mathematical task that might have “neuromorphic benefits.”
The researchers proved that neuromorphic hardware is highly energy-efficient than other systems because it can execute more random walk calculations per Joule than traditional CPUs or GPUs. Additionally, while neuromorphic chips are slower than current CPUs and GPUs, the researchers found that the difference in speed reduces when problems get bigger and more complex.
Neuromorphic hardware is still in its early stage of development, but it will eventually become easily available and smoother to program. Once this becomes a reality, the study by the research team can inspire the use of brain-inspired systems to solve mathematical problems more efficiently.
“We hope that our findings will allow random walk computational tasks to be performed far more cheaply and more energy-efficient than they are now. This, in turn, will make computing both cheaper and more climate-friendly,” Aimone added.
The latest study by Aimone and his fellow researchers concentrated more on random walk simulations such as those representing the process of diffusion. Furthermore, the team would also want to evaluate the capability of neuromorphic chips for executing more complex random walk simulations.
“Given that neuromorphic hardware continues to improve at a rapid pace and larger systems will soon be available, we expect that this advantage will continue to grow for bigger problems. There are many real-world applications that use Monte Carlo random walk models as part of their computational workload, including computational biology, materials science, financial modeling, and artificial intelligence. However, often these models are computationally expensive to run, which has huge energy, time, and financial costs,” Aimone said.
“We hypothesize that the advantage we see with neuromorphic computing will become even more pronounced with more complicated random walks, but we need to explore how to simulate more complex physics with neurons. Additionally, now that we recognize that neuromorphic hardware is well-suited for probabilistic computing applications, such as Monte Carlo random walks, we’re looking back at where the brain may use probabilistic computing in its native architecture for potential ideas at new algorithms for brain-inspired artificial intelligence,” Aimone added.