supraelectronic ANNs

(Click below images to enlarge)


Our EDA automated process takes numerically (above) processed neural networks, (implemented by inefficient digital computation), and converts to true analog supraelectronic ANNs without numerics.

Supraelectronic ANNs utilize our novel semiconductor device analog properties and processes to replicate brain neuron interactions and signal impulse response without numerical computation.

Resulting wireless supraelectronic ANNs are implemented using III-V semiconductors in our foundry with sub millimeter wave 3+ THz operation. The new adaptive ANNs mimic brain network functionality. Supraelectronic machine learning networks are of any size and complexity with unlimited SiP wireless connection fan in and fan out.

Biological neurons have tens of millions of neural connections. This is impossible with current digital ANNs and greatly limits their capabilities. Superaelectronic ANNs use wireless semiconductor neurons which approach biological neuron connection fan in and fan out.