An Intuitive Tutorial to Gaussian Processes Regression

Published:

The whole notebook can be executed at

Open In Colab

The audience of this tutorial is the one who wants to use GP but not feels comfortable using it. This happens to me after finishing reading the first two chapters of the textbook Gaussian Process for Machine Learning [1]. There is a gap between the usage of GP and feel comfortable using it due to the difficulties in understanding the theory. When I was reading the textbook and watching tutorial videos online, I can follow the majority without too many difficulties. The content kind of makes sense to me. But even when I am trying to talk to myself what GP is, the big picture is blurry. After keep trying to understand GP from various recourses, including textbooks, blog posts, and open-sourced codes, I get my understandings sorted and summarize them up from my perspective.

One thing I realized the difficulties in understanding GP is due to background varies, everyone has different knowledge. To understand GP, even to the intuitive level, needs to know multivariable Gaussian, kernel, conditional probability. If you familiar with these, start reading from III. Math. Entry or medium-level in deep learning (application level), without a solid understanding in machine learning theory, even cause more confusion in understanding GP.