Rough Sets for a Neuromorphic CMOS System

The authors have recently produced a CMOS system that autonomously rewires it’s circuitry to learn similarly to a biological brain. Biological brains learn very fast in that their synapses mysteriously ignore what is not relevant, and connect only the most important synapses to learn from an event. Conversely, man-made ‘brains’ as in, artificial intelligence systems, need to be trained to ignore all the unimportant issues - thus draining resources. Our hypothesis is to verify whether these variations of classical Rough Set algorithms can be implemented on our neuroCMOS-FPGA, and if so, under what circumstances of uncertainty, if at all, does one algorithm do better or worse than another. Herein, our next step is to 1) emulate how the neonate’s brain grows, and 2) have a rough sets system ‘allocate’ learned synthetic synapses onto the CMOS system as it grows while discerning when one learning event has similarities to other learning events; how should a rough sets system blend the building of neonatal connectome, and at the same time meld variations of synapses that have similarities?