ALBUQUERQUE, NM — With the addition of a little math, Sandia National Laboratories researchers have shown that neuromorphic computers, which synthetically replicate the brain’s logic, can solve problems more complex than those posed by artificial intelligence and can even win a place in high performance systems. computing.
A random walk diffusion model based on data from Sandia National Laboratories algorithms running on an Intel Loihi neuromorphic platform. Video courtesy of Sandia National Laboratories. Members of the media can download the video. here (4.3MB).
The findings, detailed in a recent article in the journal Electronic Natureshow that neuromorphic simulations employing the statistical method called random walks can track X-rays passing through bone and soft tissue, diseases passing through a population, information flowing through social media, and movements in financial markets. , among other uses, said Sandia, a theoretical neuroscientist. and lead investigator James Bradley Aimone.
“Basically, we’ve shown that neuromorphic hardware can deliver computational advantages relevant to many applications, not just AI that it’s obviously related to,” Aimone said. “Newly discovered applications range from radiation transport and molecular simulations to computational finance, biology modeling and particle physics.”
In optimal cases, neuromorphic computers will solve problems faster and use less power than conventional computing, he said.
The bold claims should be of interest to the high-performance computing community because finding capabilities to solve statistical problems is a growing concern, Aimone said.
“These problems are not really suitable for GPUs [graphics processing units], which is what future exascale systems will likely be based on,” Aimone said. “What’s exciting is that no one has really looked at neuromorphic computing for these kinds of applications before.”
Sandia engineer and article author Brian Franke said, “The natural randomness of the processes you list will make them inefficient when directly mapped to vector processors like GPUs in next-generation computational efforts. In the meantime, neuromorphic architectures are an intriguing and radically different alternative to particle simulation that may lead to a scalable and energy-efficient approach to solving problems that interest us.”
Franke models photon and electron radiation to understand their effects on components.
The team successfully applied neuromorphic computing algorithms to model random walks of gas molecules diffusing through a barrier, a basic chemical problem, using the 50 million-chip Loihi platform that Sandia received about a year and a half ago from Intel. Corp., Aimone said. . “We then show that our algorithm can be extended to more sophisticated diffusion processes useful in a variety of applications.”
The statements are not intended to challenge the primacy of standard computing methods used to run utilities, desktops, and phones. “However, there are areas where the combination of computing speed and lower energy costs may make neuromorphic computing the ultimately desirable option,” he said.
Showing a neuromorphic advantage, both the IBM TrueNorth and Intel Loihi neuromorphic chips observed by Sandia National Laboratories researchers were significantly more energy efficient than conventional computing hardware. The graph shows that Loihi can perform about 10 times more calculations per unit of power than a conventional processor. Power is the limiting factor: more chips can be inserted to run things in parallel, thus faster, but the same electricity bill is produced whether it’s one computer doing everything or 10,000 computers doing everything. they do the job. Image courtesy of Sandia National Laboratories. Click on the thumbnail for a high resolution image.
Unlike the difficulties of adding qubits to quantum computers, another exciting method for going beyond the limitations of conventional computing, chips containing artificial neurons are cheap and easy to install, Aimone said.
There can still be a high cost to move data in or out of the neurochip processor. “As you collect more, the system slows down and eventually won’t work at all,” said William Severa, a Sandia mathematician and author of the paper. “But we got around this by setting up a small set of neurons that did indeed calculate summary statistics, and outputting those summaries instead of the raw data.”
Severa wrote several of the algorithms for the experiment.
Like the brain, neuromorphic computing works by electrifying small pin-shaped structures, adding tiny charges emitted by surrounding sensors until a certain electrical level is reached. The pin then, like a biological neuron, emits a short burst of electricity, an action known as spiking. Unlike the metronomic regularity with which information is transmitted in conventional computers, Aimone said, artificial neurons in neuromorphic computing flicker irregularly, like biological ones in the brain, so they can take longer to transmit. information. But because the process only depletes the energies of sensors and neurons if they contribute data, it requires less energy than formal computing, which must poll all processors, whether they contribute or not. The conceptual biologically based process has another advantage: its computing and memory components exist in the same structure, whereas conventional computing consumes energy by distant transfer between these two functions. The slow reaction time of artificial neurons may initially slow down their solutions, but this factor disappears as the number of neurons increases, so more information is available in the same time period to be totaled, Aimone said.
The process begins by using a Markov chain, a mathematical construct in which, like a Monopoly board, the next outcome depends only on the current state and not on the history of all previous states. That randomness contrasts, said Sandia mathematician and paper author Darby Smith, with most linked events. For example, he said, the number of days a patient must stay in the hospital is determined, at least partially, by the length of previous stay.
Starting with Markov random basis, the researchers used Monte Carlo simulations, a fundamental computational tool, to run a series of random walks trying to cover as many paths as possible.
“Monte Carlo algorithms are a natural solution method for radiation transport problems,” Franke said. “The particles are simulated in a process that mirrors the physical process.”
The energy of each walk was recorded as a single energy spike by an artificial neuron that read the result of each walk in turn. “This neural network is more energy efficient overall than recording every moment of every walk, as ordinary computing must do. This partially explains the speed and efficiency of the neuromorphic process,” said Aimone. More chips will help the process move faster using the same amount of power, she said.
The next version of Loihi, Sandia researcher Craig Vineyard said, will increase its current chip scale of 128,000 neurons per chip to a million. Larger scale systems then combine multiple chips on one board.
“Perhaps it makes sense that a technology like Loihi could find its way into a future high-performance computing platform,” Aimone said. “This could help make HPC much more energy efficient, environmentally friendly and generally more affordable.”
Sandia National Laboratories is a multi-mission laboratory operated by National Technology and Engineering Solutions of Sandia LLC, a wholly owned subsidiary of Honeywell International Inc., for the National Nuclear Security Administration of the US Department of Energy. Sandia Labs has significant responsibilities of research and development in nuclear deterrence, global security, defense, energy technologies, and economic competitiveness, with principal facilities in Albuquerque, New Mexico, and Livermore, California.
Sandia Media Contact: Neal Singer, firstname.lastname@example.org
Does not apply
Advantages of neuromorphic scaling for energy efficient random walk calculations
Article Publication Date