I think that's a great idea. At this point in time, there is so much stuff going on the CS/AI/ML domain that translating research papers into something people can read without incalculable suffering might be as important and valuable as research itself. After all, many of those papers concern things that almost anyone with access to a computer can actually implement and use.
PS: Bonus points for anything that diverges from the current mainstream (like SVMs or boosting).
Precisely! I highly encourage checking out the slide-deck for a graphical representation.
For every node in every other layer, I colocate the edge on the same machine. In this way, when a group of, say, 10 nodes in layer 1 are each sending a weighted message to a single node in layer 2... they can pre-combine their messages (weighted sum) and send only that value over the network. This happens for every node in the second layer, reducing network i/o (this is the first optimization).
perhaps instead of discounting his experience... it would be better to take his advice