The Learning Rate
This video presents the same text shown beside it, spoken and on screen. It adds nothing the text does not say.
State
The stride of descent is the learning rate — and the anchor rule of modern training, each weight moving by stride times error times its own input, is the perceptron's nudge with a graded error.
Show
The nudging rule was gradient descent wearing mittens; the smooth bend supplied the grading.
Watch for
One procedure spans the unit — the continuity is the insight, not a coincidence.
Builds on
Unlocks
- Nothing yet depends on this.