"Homework: New Research Suggests It May Be an Unnecessary Evil" by Alfie Kohn

[DetlefK]

12. positive and negative feedback

If the result delivered by the NN is equal to the demanded outcome, then there's a positive feedback and the NN is modified in a way to get a similar positive outcome next time.
If the result isn't equal to the demanded outcome, then there's a negative feedback and the configurations that led to the result are weakened.
(It's a little bit more complicated in fuzzy-logic. The neurons do not operate on true/false, but have a continuous scale from true to false.)

In practice, it's important to know when to stop educating a neural network and keep its configuration from changing again:
If you don't give them enough experience, they will think too broadly and might react wrong.
If you give them too much experience, they won't be able to think "outside of the box" and might react wrong as well.
IIRC the human brain enters this stage in the mid-twenties and knowledge/abilities/character become more and more fixed over time.



And why homework? Because the brain of a child is learning all the time. It's just a matter of priorities WHAT to give it for learning.