Hypotheses
I maintain a list of loose hypotheses, e.g. predictions / expectations, and track (solely empirical) evidence as I find it. It is my belief (hypothesis!) that empirical evidence is more important than theoretical ideas about performance or behavior of networks.
-
.#soc-efficiency
-
small-kernels-depth-efficiency
Smaller kernels over multiple layers outperform shallower stacks with larger kernels
-
soc-efficiency
Separating concerns amongst substantially independent agents is efficient