You probably do want to include library code for fast random projections based on random sign flipping and the Walsh Hadamard transform, even if it is just to allow people to experiment.
https://discourse.numenta.org/t/non-linearity-sharing-in-deep-neural-networks-a-flaw/6033