liveBook · Manning

Gaussian process

As explained in Grokking Bayes

A Gaussian process is a probability distribution over functions such that every finite collection of function values has a joint Gaussian distribution.

A Gaussian process is a probability distribution over functions: every finite collection of function values has a joint Gaussian distribution. Intuitively, it extends Bayesian linear regression beyond a single straight-line relationship. section 3.16 section 3.18

Linear regression assumes that labels follow a linear relationship with features. More complex data may call for higher-degree polynomials; pushing that flexibility to the extreme gives a Gaussian process, interpretable as an infinite-degree polynomial model. section 3.16

The key progression is:

  • Linear regression models a first-degree polynomial relationship. section 3.16
  • Higher-degree polynomial models represent increasingly complex trends. section 3.16
  • Gaussian processes generalize Bayesian linear regression to infinite-degree polynomials while retaining distributional uncertainty. section 3.18

How Grokking Bayes covers it

How deep it goes
Interprets Gaussian processes as infinite-degree polynomial models but defers detailed treatment and applications.
Examples
Uses linear relationships, higher-degree polynomials, and complex real-world trends as comparison points.
What you can do after
Evaluate/understand Gaussian processes as flexible generalizations of Bayesian linear regression for complex trends.
How it connects
Connects Gaussian processes to linear regression, Bayesian linear regression, polynomial models, and predictive uncertainty.

Also explained in

Compare how 4 books explain Gaussian process →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Grokking Bayes. The section links point to the parts of the book it draws on.