tinyML Talks: The new Neuromorphic Analog Signal Processor (NASP) concept and technology platform 

"The new Neuromorphic Analog Signal Processor (NASP) concept and technology platform"
Alexander Timofeev
Founder and Chief Executive Officer
Polyn.ai
One of the major advantages of the neural network is its ability to do things in parallel. The best way to implement this parallelism is neuromorphic computing and new architectures that mimic the systems in the human brain and sensory nervous system to perform with high accuracy and ultra-low power consumption.
Neuromorphic Analog Signal Processing (NASP) technology is based on several major advances. First, it is a neural network operation (inference) in the analog neuro core with fixed weights and structure, performed separately from its training in a digital framework.
The NASP T-compiler allows transformation of almost any neural network into an internal representation that is transferred 1:1 into the analog neuro core on a standard CMOS chip.
The NASP chip is a technology breakthrough aiming to add intelligence to various sensors. It contains artificial neurons (nodes performing computations) and axons (connections with weights between the nodes) implemented using analog circuitry elements.
The NASP hybrid core concept combines a fixed neuromorphic analog core (with ultra-low power consumption and low latency) taking a major computational load of data pre-processing, with a fully flexible digital part for a minor load of final classification.