V1 Pattern Learning - Version 2

Learns 20 visual patterns presented randomly using at most 40 neurons. We can see : • The system slowly converging to 20 neurons (the system isn't aware there are 20 patterns to discover) • The normalized residual energy going from 100% (not covered) down to 0% (fully covered) • The ON/OFF activation histogram reporting less and less uncovered signal (at the end, only the noise remains) • V1 quality values all converging to 100% (meaning a good coverage of both the pattern and its amplitude) • The bottom left Pattern ID image shows green everywhere, meaning there's no conflict among neurons : each neuron is covering a different pattern