Researchers Devise ‘Honey Memristor’ for Neuromorphic Chips
Table of Contents
Check the date — this appears not to be an April Fool’s joke. Research engineers from WSU’s School of Engineering and Computer Science, Feng Zhao and Brandon Sueoka, have devised a ‘honey memristor,’ which they hope to use en masse as a key component of neuromorphic hardware. Their report appears in Journal of Physics D.
The process is deceptively simple. First, the researchers processed a sample of honey into a solid. Then they compressed it between two metal electrodes. The design is known as a ‘memristor,’ a portmanteau of ‘memory’ and ‘resistor.’
What makes this invention a memristor, instead of two sticky pieces of metal, is a physical property of the honey itself. Specifically, memristor materials have a voltage-dependent resistance. Their structure changes, reversibly, creating a kind of physical history of electrical activity across the memristor. Honey fits that description, since it changes reversibly between a viscous liquid and a crystalline solid. For that reason, honey can serve as the physical history. That’s what makes these honey memristors a candidate for use in neuromorphic systems.
Buzzwords
The researchers’ lofty hope for their honey memristor is that as neuromorphic hardware, it will be able to perform at a level rivaling or even outstripping “traditional” von Neumann architectures. But they also characterize their invention as renewable, biodegradable, non-toxic, low-power, and even antibacterial. “Honey does not spoil,” said corresponding author Feng Zhao. “It has a very low moisture concentration, so bacteria cannot survive in it. This means these computer chips will be very stable and reliable for a very long time.”
“When we want to dispose of devices using computer chips made of honey, we can easily dissolve them in water,” he added. “Because of these special properties, honey is very useful for creating renewable and biodegradable neuromorphic systems.”
There are some hurdles that they’ll have to overcome. The most important might be the electrical and thermal limits of the device. Zhao and Sueoka offer the honey memristor as a component for neuromorphic systems because they operate at low power. One of the big selling points of neuromorphic systems is that they run at just a few dozen watts, as opposed to the hundreds of watts a consumer or server CPU can draw. The honey memristor can only handle twenty watts at most, and it has corresponding thermal boundaries.
Neuromorphic computing isn’t suited for all tasks. It’s unlikely that we’ll see neuromorphic processors in desktop hardware anytime soon. Instead, neuromorphic architectures like Intel’s Loihi lend themselves to big-data research tasks.
In addition to his team’s work with honey, Zhao hopes to continue exploring other renewable and biological solutions for use in neuromorphic computing. One potential target: the proteins and sugars found in aloe vera leaves.
Feature image by Jon Sullivan, Wikipedia
