Comments by "Chrysippus" (@4.0.4) on "Backpropagation, step-by-step | DL3" video.

  1. At 9:50 the notion of "mini-batches" is mentioned, and I can't avoid thinking, why not smaller batches? Or larger? How is the optimal size for those determined? For that matter, I also still don't know how to decide number of hidden layers/neurons. There's a proper way, right?
    1