Putting a little Topspin on neuromorphic computing, digital platforms and services

Collaborating universities in Sweden and Japan have made a breakthrough in the ongoing quest to make computers more energy efficient by making them function more like the human brain. Scientists from the University of Gothenburg and Tohoku University, which is located in the city of Sendai, 380 kilometers north of Tokyo, have done so under the auspices of the Topotronic Multi-Dimensional Spin Hall Nano-Oscillator Networks research project, which it is usually rather more nimbly known as “Topspin” (as in tennis or snooker).

Topspin’s purpose is to be a catalyst to help create more efficient technologies in sectors such as mobile phones, satellites and autonomous vehicles. In this instance, the result of academic collaboration is demonstrating, for the first time, that it is possible to unite oscillators and memristors, merging them into a single unit that combines memory and calculation functions.

Recently, research interest in using coupled oscillator networks to improve computation has increased dramatically due to the belief that oscillator-based systems can be developed faster and better than traditional digital circuits. For its part, a memristor limits or regulates the flow of electrical current in a circuit and remembers the amount of charge that has previously circulated through it. In other words, they are not volatile and that is very important because they retain memory without power.

Memristor-controlled arrays of oscillators have been found to come close to emulating neural networks in the human brain that somehow seem to work on oscillatory signals. This is a seductive notion and work on artificial neutrons and synapses has been going on, with considerable success, for some years. Topspin’s joint Swedish/Japanese research shows that oscillators and oscillating circuits can perform complex calculations in a way that mimics how human nerve and memory cells appear to do the same thing.

Quoted in the academic journal ‘Nature Materials’, Johan Åkerman, Professor of Applied Spintronics at the Department of Physics at the University of Gothenburg, says: “Computers are now incredibly good at completing advanced cognitive tasks, such as language and word recognition. images or display superhuman chess skills, thanks in large part to artificial intelligence. At the same time, the human brain is still unmatched in its ability to perform tasks effectively and energy efficiently. Finding new ways to perform calculations that resemble energy-efficient processes in the brain has been a major focus of research for decades.”

He adds: “Cognitive tasks… require significant computing power, and mobile applications, in particular, such as drones and satellites, require energy-efficient solutions. This is an important advance because we show that it is possible to combine a memory function with a calculation function in the same component. These components function more like the brain’s energy-efficient neural networks, allowing them to become important building blocks in future, more brain-like computers.”

When it comes to better and more energy efficient mobile phones, Professor Åkerman believes that the new research will lead to new functionality in the devices. He uses the example of digital assistants like Siri where currently all processing has to be done by remote servers because such heavy processing work on a mobile phone is very energy inefficient. However, if the components could be made small enough, many hundreds of them could fit in a mobile phone, allowing for energy-efficient local processing and obviating the need for power-hungry servers. Under laboratory conditions, research scientists have been able to produce components so microscopically small that they are “less than the size of a single bacterium.”

By the way, spintronics, or spin electronics, is a technology that exploits the intrinsic spin properties of the electron. An electron can exist in one of two spin states: spin up and spin down, that is, it can spin clockwise or counterclockwise with constant frequency around its axis. This property can be used to represent a 0 or a 1 in logical operations.

Leave a Reply

Your email address will not be published.