Loading slide

Why Shallow Networks Failed

  1. 01Why Shallow Networks Failed
  2. 02Memory card
  3. 03What "deep" actually means
  4. 04The vanishing gradient
  5. 05A small step inside each neuron
  6. 06Why bother bending the number?
  7. 07The bend they chose
  8. 08How the flat ends starve the signal
  9. 09The absurdly simple fix
  10. 10Take a breath
  11. 11Other fixes
  12. 12Depth vs. width
  13. 13Memory card
  14. 14The research gap
  15. 15Reinforce your understanding
  16. 16Question: The vanishing gradient
  17. 17Question: Why does a simple fix work?
  18. 18Question: Why does depth matter?
  19. 19Quiz: answer
  20. 20The one idea to keep
  21. 21The gradient problem was solved. What was still missing was scale.
  22. 22Want to go deeper?
9 / 21
BackNext

The absurdly simple fix

The fix is almost comically simple. So simple it seemed, at first, like it couldn't possibly be the answer.

Drop the S-curve. Replace it with one plain rule: if the number is positive, pass it straight through, unchanged. If it is negative, output zero. That is the whole thing, and it has a name, .

Now, you might object: "pass it through unchanged" sounds like doing nothing, like multiplying by 1. And did we not just say a chain of plain multiplies collapses into one? So how is this any different?

Here is the trick, and it is the word if. ReLU is not one rule, it is two rules with a corner between them. Flat at zero for every negative input, then it turns and climbs straight up for every positive one. Draw it and you get an elbow. That corner is a bend, a real one, just like the sigmoid's curve was a bend. So ReLU does not collapse. Stack many of them and you get many corners, an intricate folded shape no single multiply could make.

But look at the difference from sigmoid. The sigmoid bent by going flat at its ends, and flat is where the signal died, multiplied down by 0.2 again and again. ReLU's bend never flattens on the live side. For a positive number the line stays at full steepness, so the backward signal passes through at full strength, times one, not times a fraction. It either continues undimmed or, on the zero side, stops cleanly.

So ReLU threads the needle. It keeps a bend, so depth still means something. But it throws away the flat parts that were strangling the signal. One tiny change to the step inside each neuron, and deep networks became trainable at last.

Citations(2)↓
  1. 1. proceedings.mlr.press
  2. 2. ibm.com
Citations(2)↓
  1. 1. proceedings.mlr.press
  2. 2. ibm.com