Learning Rate
The size of each gradient descent step: too large overshoots and bounces around, too small crawls and wastes time, just right converges steadily.
When to use
Use it when asked what goes wrong in training: an oversized step overshoots the minimum of the loss, an undersized one is very slow.
See Lec 8.