Loading slide

A Network That Learns

  1. 01A Network That Learns
  2. 02Stacking neurons
  3. 03Quick recall
  4. 04What stacking does
  5. 05Hidden layers
  6. 06The problem of credit
  7. 07Backpropagation
  8. 08Sharing the blame
  9. 09The search problem
  10. 10Fog on a hill
  11. 11Gradient descent
  12. 12Just to be clear
  13. 13What the hidden layer learns
  14. 14What the hidden layer discovers
  15. 15Did it actually learn?
  16. 16Generalization
  17. 17Limits in the 1990s
  18. 18Reinforce your understanding
  19. 19Question: Backpropagation
  20. 20Question: What the hidden layer learns
  21. 21Question: Memorizing vs. generalizing
  22. 22Quiz: answer
  23. 23The method existed. What was missing was scale.
  24. 24Want to go deeper?
9 / 23
BackNext

The search problem

The other half is knowing what to actually do with that information.

You can't just "fix" the weights. There's no single correct value for any of them. A network might have millions of weights, each one a dial that can be set to any number.

Backpropagation tells you which direction each dial should turn. But you still have to decide how far to turn it, when to stop, and how to do this across millions of examples without losing your way.

Training is a search. The question is how to search well.

Citations