A study led by a James Cook University scientist shows computer chips based on biological brains may be the solution to a looming bottleneck in computing speed.
Dr Mostafa Rahimi Azghadi is a senior lecturer in electronics and computer engineering at JCU. He led an international team that examined the cutting edge of computer design – with the results recently published as a cover article in the journal Advanced Intelligent Systems.
Dr Rahimi Azghadi said the ever-increasing processing power demands of computers cannot continue to be fulfilled indefinitely unless there is a paradigm shift in the field.
“Conventional computing involves numerous interactions between the slow memory and fast processors and the data movement between memory and processor produces a bottleneck,” he said.
Dr Rahimi Azghadi said this significantly affects computer speed, especially when a large amount of data needs to be processed in a short time and makes for very high power consumption.
He said the increasing importance of big data and a need for low-power and high-speed processing in everyday life demands a shift to computers with advanced capabilities but low power consumption.
“Neuromorphic computing, which takes inspiration from the efficient and low-power computing capabilities of the brain, may provide an answer,” he said.
Dr Rahimi Azghadi said computer designers can reproduce almost any behavioural characteristics of neurons or synapses, if they have been mapped through a mathematical model.
“Then it’s a case of routing together transistors and capacitors to follow these mathematical rules,” he said.
In an earlier study, Dr Rahimi Azghadi and his collaborators developed digital computer hardware that learns to recognise simple digits, by implementing a physiologically measured and computationally modelled brain learning principle.
He said neuromorphic engineering can be broadened to any biological process, but is currently mainly focused on learning and cognition in the brain.
“Researchers have a strong idea of the architecture of biological cells in sensory systems such as the retina and, to some extent, the brain.
“What is lacking is a unified account of the computation that takes place within biological sensors and how they process, filter, and store information before transmitting them to the brain.”
He said researchers at JCU are currently working on using emerging hardware technologies to clear the way for the implementation of neuromorphic systems and to accelerate commonplace machine-learning algorithms.