Neuromorphic Computing

Neuromorphic Computing – The Next Gen AI
: We’ve all heard of artificial intelligence and Neural Networks. They can help to find out patterns in vast volumes of data. But in order to understand things of greater complexity like the human brain and cognition we need computers that think like our brains!

This is what neuromorphic  computation does. In this type of computing, the computational architecture of the  humanbrain is emulated. Traditional chips work on Von Neuman architecture which has a separate  CPU, memory unit and data paths. On the contrary, our brain consists of a vast network of billions of neurons that are connected to each other by synapses. Each  neuron acts as an independent processing and storage unit and can fire independently. Neuromorphic computation tries to mimic this system by implementing the neuro – biologicalarchitecture onto electronic circuits. This is done by implementing Spiking Neuron Models onto electronic devices like FPGAs and GPUs.

The specialties of devices like GPUs and FPGAs are their parallel processing capabilities due to the GPUs multi cores and the FPGAs parallel architecture. Implementation of this type of computing on hardware can also be done using technologies like memristers, spintronic memories, nanocrystals , # nanowires , conducting  polymers , etc. Companies have also developed dedicated chips for this like Intel’s ‘Loihi’ Chips, IBM’s TrueNorth Chip and Brainchip’s Akida NSoC.
This  technology can be used to train machines to recognize patterns with far fewer inputs than that required to train digital neural networks. It can also be used for brain simulations and to understand brain abnormalities like  alzheimer and ALS as well as help in  geneticengineering

This is a very vast field that lies on the intersection of computerscienceneurobiology ,electronics and mathematics but it has the potential to revolutionize traditional computation.

-Author Rishona Daniels
-Creatives: Anupriya Dasgupta