What scientists do for a decades?
In an effort to design machine learning systems that are quicker and more energy-efficient, researchers have been looking into ways to replicate the flexible computing capabilities of biological neurons. Memristors, electronic devices that can store a value by changing their conductance and then use that value for in-memory processing, are one promising strategy.
How the study is going by the researchers?
By enabling the electronic memristors in the study to respond to visual as well as electrical inputs, the researchers increased their functionality. Due to this, distinct feedforward and feedback channels could exist within the network at the same time. The development enabled the researchers to develop winner-take-all neural networks, computational learning programmes that have the ability to resolve challenging machine learning issues, including unsupervised learning in clustering and combinatorial optimisation issues.
Why artificial neural networks are the highly exciting development?
This is a very exciting development. Our research has developed a revolutionary idea that outperforms the fixed feedforward operation currently used in artificial neural networks. These latest proof-of-principle results show an important scientific progress in the broader domains of neuromorphic engineering and algorithms, enabling us to better replicate and grasp the brain, aside from the possible applications in AI technology.
We have found more exploratory information than system-level demos. Although we intend to develop this idea further in the future, we are certain that our current proof-of-concept outcomes show a significant scientific interest in the broader domains of neuromorphic engineering and help us better understand and replicate the brain.