Stop Guessing: Backpropagation to Code (Part 4)

Added: 2026-05-05

[00:00:00]So, how should we change our W's to decrease our cost?

[00:00:04]We can now compute DJDW, [music] which tells us which way is uphill in our nine-dimensional optimization space.

[00:00:12]If we move this way [music] by adding a scalar times our derivative to all of our weights, our cost will increase.

[00:00:17][music] And if we do the opposite, subtract our gradient from our weights, we will move downhill and reduce our cost.

[00:00:25]This simple step downhill is the core of gradient descent >> [music] >> and a key part