Many researchers around the world have been attempting to construct brain-inspired computer systems for the last decade or so. Currently, the majority of these systems are utilised to run deep learning algorithms and other artificial intelligence (AI) technologies. The technology is also known as neuromorphic computing tools during the last decade or so.
Recently, a research team has undertaken research on the ability of neuromorphic architectures to execute a novel type of computation, namely random walk computations. These are calculations involving a series of random steps in mathematical space. The researchers' findings are highly relevant to the Neuromorphic Computing Market. It indicates that neuromorphic architectures may be well-suited for implementing these calculations. Thus, extending beyond machine learning applications.
Neuromorphic hardware is currently in its early phases of development. However, it is expected that it will become more widely available and easier to programme over time. Once this occurs, the latest work conducted by this group of academics may motivate the adoption of brain-inspired systems to solve mathematical problems more effectively.
Researchers anticipate that the discoveries will enable random walk computing tasks to be executed significantly more cheaply and efficiently than they are currently. As a result, computing will become both cheaper and more environmentally benign.
Further, the team demonstrated that neuromorphic hardware is more energy-efficient than other systems because it can perform more random walk calculations per Joule than traditional CPUs and GPUs. It is true that neuromorphic devices are currently slower than present CPUs and GPUs. But the scientists discovered that this speed difference lessens as issues grow larger and more complicated.
Primarily concentrated on simple random walk simulations, such as those modelling the diffusion process. However, the team would like to investigate the potential of neuromorphic circuits for executing more complex random walk simulations in the future.
The group believes that the benefit found with neuromorphic computing would become much more evident with more intricate random walks. However, the team needs to research how to model more complex physics with neurons. Additionally, now they recognise that neuromorphic hardware is well-suited for probabilistic computing applications like Monte Carlo random walks. Hence, they are looking back at how the brain may use probabilistic computing in its native architecture for potential ideas. This will help develop novel algorithms for brain-inspired AI.