Numerical Differentiation to Find the Gradient
from May be duplicated
Numerical Differentiation to Find the Gradient
https://gyazo.com/8cba7761543fa4a71b925f12f791b94c
Vertical axis is goodness, horizontal axis is search space.
Not sure which way to go if there is only one observation
Given two observations, we can compare them to know "Better Direction".
This is gradient by numerical differentiation.
The figure determines that the further to the right, the larger
In the figure, the search space is only left and right because it is 1-dimensional, but in general, the search space is high-dimensional
Proceeding in the direction of the gradient yields better results.
This is the least-descent method.
---
This page is auto-translated from /nishio/数値微分で勾配を求めるたとえ using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.