A team of researchers from the University of Sydney and Japan’s National Institute for Material Science have demonstrated that they can utilize a random network of nanowires to mimic the structure as well as the dynamics of the brain to solve simple tasks involving processing.
Deep neural networks already resemble one aspect of the brain that is a highly interconnected network of neurons. However, artificial neurons behave in a very different manner as compared to biological ones since they only perform computations. In the human brain, neurons can also remember their previous activity, which dramatically influences their future behavior. The in-built memory is an essential aspect of how the brain processes information. A significant strand in neuromorphic engineering focuses on trying to recreate this functionality.
Researchers have also found memristive properties at the junctions where silver nanowires overlap with each other. They also have the added benefit that they can self-assemble into complex networks, unlike those found in the brain.
The researchers created a random network of nanowires 10 micrometers long and no thicker than 500 nanometers and. The network was then subjected to electrical stimulation.
As current passed through the network, the memristive junctions switched on and off, altering the signal’s path. However, the pattern of switching varied considerably depending on the strength of the input signal. To test if these dynamics could be used for information processing, the researchers created a network simulation. They tried to teach it how to carry out a simple signal processing task, converting one waveform into another.
To perform this, they adjusted the amplitude and frequency of the input to check if this would impact the performance. The team found that the network did best when the strength of the signal was right on the verge of pushing the network into a chaotic state. This is interesting because it’s been speculated that the human brain operates in this regime too.
Another intriguing finding the researchers made was that the “edge-of-chaos” state was most potent when trying to convert between the most different waveforms. This suggests that while the approach may not be efficient for simple tasks, it is particularly suitable for more complex ones.
However, there is still a long way before these nanowire networks are anywhere close to matching the power of the human brain. The researchers demonstrated a very simple task, and it has been shown in a simulation of the network rather than the real network. The results, however, are promising as they did give weight to the argument that nanowire networks could be a tentative avenue for recreating the powerful and energy-efficient processing of the brain.