New Delhi (ISW). New dimensions are being created all over the world in the field of Artificial Intelligence (AI). Working in this direction, Indian researchers have got a big achievement.
Researchers at the Indian Institute of Technology (IIT), Delhi in their latest study have developed an artificial neuron inspired by the brain, which could be useful in building an accurate and efficient neuromorphic artificial intelligence system.
Neuromorphic computing refers to the area of technology where engineers try to build intelligent machines inspired by the workings of the human brain.
Neurons and synapses are believed to be the most important building blocks of intelligence within the brain.
The researchers say that due to the interdisciplinary nature of this research, it connects different areas such as AI, neuromorphic hardware, and nanoelectronics.
Developed under the leadership of Professor Manan Suri, Department of Electrical Engineering, IIT Delhi, this new 'spiking' neuron model has been named DEXAT (Double Exponential Adaptive Threshold Neuron).
This research is important, as it could be used to build accurate, fast, and energy-efficient neuromorphic AI systems for real applications such as speech recognition.
Features of the newly developed artificial neuron include a new multi-time scale spiking neuron model for neuromorphic processing called DEXAT.
In this study, efficient hardware demonstration of a new neuron model based on an emerging nanoelectronic device and successful hardware demonstration of real-time Spatio-temporal neuromorphic processing is demonstrated.
Professor Manan Suri said, 'We are conducting extensive research on all aspects of semiconductor memory technology and its emerging applications with academic and industrial partnerships.
Over the years we have successfully demonstrated many innovative uses of memory technology beyond simple storage.
Semiconductor memory has been used efficiently for applications such as in-memory-computing, neuromorphic-computing, edge-AI, sensing, and hardware security.
This research uses the analog properties of nanoscale oxide-based memory devices specifically to fabricate adaptive spiking neurons.
High performance with fewer neurons: This research demonstrates a flexible spiking neuron model with more precise, faster convergence and efficient hardware implementation than existing modern adaptive threshold spiking neurons.
The new solution achieves high performance with fewer neurons. Related benefits have been shown on several data sets. Classification accuracy of 91 percent has been achieved on the Google Spoken Command dataset. A patent has also been filed on this research work.
Good results were seen: This research has been published in Nature Communications. Apart from Professor Manan Suri, two other researchers, Ahmed Shaban and Sai Sukruth Bejugam are involved in this research work.
Researchers have successfully demonstrated a hybrid nano device-based hardware.
The proposed nano-device neuromorphic network achieved 94% accuracy even with high device variability, indicating its robustness.
IIT Delhi has also tweeted this success of Manan Suri. IIT director V Ramgopal Rao has also congratulated him on his success by tweeting.
Posts
