So, apparently "sharp" cos similarity can be as good as convolutions but with 10-100x less compute?
CleverLibre Social is an inclusive social instance for open discussion, learning, and community. All cultures welcome. Hate speech and harassment strictly forbidden.
