The Matrix Cookbook

Learn / The Matrix Cookbook

08 - Gaussians

Advanced formula selection, assumption checking, and exact application for Gaussians, book pages 40-45.

Learning path

0%

0 of 3 sections marked complete · about 25 minutes

Learning objectives

What you will be able to explain

  • Affine Gaussian transformations
  • Product of Gaussian densities
  • One-dimensional precision fusion
  • Gaussian conditioning blocks
  • Gaussian moments
  • Gaussian comparison structure

Section 01

Affine Gaussian transformations to One-dimensional precision fusion

01

Affine Gaussian transformations

Gaussian families are closed under affine maps; constants affect means, not covariance.

Guided checkpoint

Let xN(mu,Sigma)x\sim N(mu,Sigma) and y=Ax+by=Ax+b.

Source: Sec. 8.1, pp. 40-42

02

Product of Gaussian densities

Completing the square adds quadratic precision terms; the mean is precision weighted.

Guided checkpoint

The product of two Gaussian densities in the same variable is proportional to what?

Source: Sec. 8.1, pp. 40-42

03

One-dimensional precision fusion

Precision-weighted mean is (0/4+3/1)/(1/4+1)=3/1.25=2.4(0/4+3/1)/(1/4+1)=3/1.25=2.4.

Guided checkpoint

Combine N(0,4)N(0,4) and N(3,1)N(3,1) as a product density. What is the resulting mean?

Source: Sec. 8.1, pp. 40-42

Section 02

Gaussian conditioning blocks to Gaussian comparison structure

01

Gaussian conditioning blocks

Conditioning updates the mean using the observed residual and reduces uncertainty by a Schur-complement term.

Guided checkpoint

For jointly Gaussian [x;y][x;y] with covariance blocks, match the conditional quantity.

Source: Sec. 8.1, pp. 40-42

02

Gaussian moments

Isserlis/Wick pairing determines even moments; joint Gaussian zero covariance implies independence.

Guided checkpoint

For centered Gaussian variables, apply the chapter's moment rules.

Source: Sec. 8.2, pp. 42-44

03

Gaussian comparison structure

Gaussian comparisons combine volume, shape, and displacement.

Guided checkpoint

Which ingredients determine divergence-like comparisons between nonsingular Gaussians?

Source: Sec. 8.3, p. 44

Section 03

Mixture of Gaussians to Compute a mixture mean

01

Mixture of Gaussians

A mixture averages densities, not parameters; between-component spread contributes to total covariance.

Guided checkpoint

For weights pikpi_k summing to one.

Source: Sec. 8.4, pp. 44-45

02

Compute a mixture mean

.25(2)+.75(4)=.5+3=2.5.25(-2)+.75(4)=-.5+3=2.5.

Guided checkpoint

A two-component scalar mixture has weights .25 and .75 and means -2 and 4. Compute its mean.

Source: Sec. 8.4, pp. 44-45

Knowledge check

Turn understanding into recall.

The quiz now follows the same concepts in scored form. You can return to this lesson from the quiz whenever a gap appears.