i like how it suggests that backprop may really be the best algorithm for training neural networks
Conversation
Replying to
This is confirmation that there is not a simpler yet more efficient way of doing it. There might be "better" training algorithms (harder to find by both humans and search methods)


