Follow
I wonder how many people realize that neural nets are a special case of cellular automata, and that by studying how to optimize their computation you can achieve a more general solution
@pee_zombie what would happen if you knew how many
@Coke id know how many
neural nets are non-grid CA with locally-variable neighborhoods (edge weights)
CA evaluations can be made hyperefficient through adjacency matrix compositions
by leveraging the equivalence posited earlier, you can optimize model evaluation
training is still expensive tho