The nanowire network learns and remembers like the human brain

An international team led by scientists from the University of Sydney has shown that nanowire networks can exhibit both short- and long-term memory, similar to the human brain.

The research is published today in the journal Science Advances, Dr. Led by Alon Loeffler, who earned a Ph.D. in the School of Physics, with collaborators in Japan.

“In this research we found that higher-order cognitive function, which we normally associate with the human brain, can be simulated in non-biological hardware,” said Dr. Loeffler said.

“This work builds on our previous research in which we showed how nanotechnology can be used to create brain-inspired electrical devices with neural network-like circuitry and synapse-like signaling.

“Our current work paves the way for replicating brain-like learning and memory in non-biological hardware systems and suggests that the underlying nature of brain-like intelligence may be physical.”

A nanowire network is a type of nanotechnology that is typically made of tiny, highly conductive silver wires that are invisible to the naked eye, covered in a plastic material, and interspersed like a mesh. Wires mimic aspects of the physical structure of the human brain network.

Advances in nanowire networks could improve many real-world applications, such as robotics or sensor devices that require rapid decision-making in unpredictable environments.

“This nanowire network is like a synthetic neural network because the nanowires act like neurons, and the places where they connect to each other are similar to synapses,” said senior author Professor Zdenka Kuncic from the School of Physics.

“Instead of implementing some kind of machine learning task, in this study Dr. Loeffler has actually taken it a step further and tried to show that the nanowire network performs some kind of cognitive function.”

To test the nanowire network’s capabilities, the researchers gave it a test similar to a common memory task used in human psychology experiments, called the N-Back task.

For a person, an N-Back task might involve remembering a specific picture of a cat from a series of cat images presented in sequence. An n-back score of 7, the average for people, indicates that the person can recognize the same image that appeared seven steps back.

When applied to a nanowire network, the researchers found that it could ‘remember’ the desired endpoint in an electric circuit seven steps back, meaning a score of 7 in the n-back test.

“What we did here is manipulate the voltage at the end electrodes to force the pathways to change instead of just letting the network do its thing. We forced the pathway to go where we wanted it to go,” Dr. Loeffler said.

“When we implemented it, its memory had a very high accuracy and didn’t really degrade over time, which suggests that we’ve found a way to strengthen the pathways to push them to where we look, and then the network remembers it.

“Neuroscientists think that this is how the brain works, that certain synaptic connections are strengthened while others are weakened, and that’s how we preferentially remember certain things, how we learn, etc.”

The researchers said that when the nanowire network is continuously reinforced, it reaches a point where it no longer needs reinforcement as the information is integrated into the memory.

“This is like the difference between long-term memory and short-term memory in our brains,” Professor Kuncic said.

“If we want to remember something for a long time, we really need to keep training our brains to consolidate it, otherwise it kind of fades away over time.

“One work shows that a nanowire network can store seven items in memory at significantly higher than chance levels without reinforcement training and with near-perfect accuracy with reinforcement training.”