What's the status of Neuromorphic chip development? Is it proven? Is it getting mature to production or application? Or is it still early research and we'll have to wait 3-5 years before seeing any real usage?
The wonderful fact about the human brain is that its unique computational capabilities and its inherent ability to adapt and learn. Human brain can execute multiple advanced computing tasks that includes learning, recognition, and cognition. All these tasks are performed using extremely low power consumption (< 50 W) and low frequency of neuronal spiking that is attributed to the highly-parallel and the event-driven scheme of computation, where energy is used only when and where it is needed for processing the information. Mimicking the human brain is a daunting task that requires the replication of the time-dependent plasticity of synapses and the achievement of the high connectivity in biological neuron networks, where the ratio between synapses and neurons is around 10^4, which is excessively high. To mimics the neuro-biological architectures, Neuromorphic (Brain-like) computing concept was originally designed as the hardware for implementing models of neural systems that was further extended to the computing systems that can run bio-inspired computing models such as, neural networks and deep learning networks etc. The end objective of creating architectures for neuromorphic computing is to create an electronic brain. A tremendous progress has been made on building the Architectures for neuromorphic computing. There is a need significant amount work requires to be done for designing software for the implementation and leveraging all these architectures to solve real life problems.
Recent neuromorphic processor test chips & software
- TrueNorth chip by IBM
- SpiNNaker hybrid CPU/Neuron project – University of Manchester
- FACETS – University of Heidelberg
- Darwin: A neuromorphic hardware co-processor
- DANNA: A neuromorphic software ecosystem
- FPGA-based neuromorphic-like embedded system
- Adaptive Neuromorphic Architecture (ANA)
- Spiking Neural Networks (SNN) for Versatile Applications (SNAVA)
- MoNETA: HP and U of Boston's Approach