Course contentsShow
← Machine Learning and Deep Learning
Lesson 18 of 3,5381. Mathematical Foundations for Machine LearningFree lesson

Eigendecomposition of Matrices

Master decomposing matrices as A=QΛQ⁻¹ and understand when this decomposition exists.

Eigendecomposition of Matrices

What you'll learn: How to break down certain matrices into simpler building blocks using eigenvalues and eigenvectors.

The Core Idea

Eigendecomposition is like disassembling a machine into its fundamental parts to understand how it works. Given a square matrix A, eigendecomposition expresses it as:

A = QΛQ⁻¹

where:

  • Q is a matrix whose columns are the eigenvectors of A
  • Λ (lambda) is a diagonal matrix containing the eigenvalues on its diagonal
  • Q⁻¹ is the inverse of Q

Think of it this way: the matrix A transforms vectors in complex ways, but eigendecomposition reveals that A simply stretches space along specific directions (the eigenvectors) by specific amounts (the eigenvalues).

When Does Eigendecomposition Exist?

Not every matrix can be eigendecomposed! For eigendecomposition to work:

  1. A must be a square matrix (same number of rows and columns)
  2. A must have enough linearly independent eigenvectors to form Q

Special case: Symmetric matrices (where A = Aᵀ) are always eigendecomposable, and their eigenvectors are orthogonal (perpendicular to each other). This makes them especially nice to work with!

Why It Matters

Eigendecomposition lets you understand what a matrix "really does" by separating the directions of transformation (eigenvectors) from the amount of stretching (eigenvalues). This decomposition also makes certain matrix operations much faster to compute.

Key Takeaway: Eigendecomposition breaks a matrix A into A = QΛQ⁻¹, revealing its fundamental behavior through eigenvectors and eigenvalues—but only works when A has enough independent eigenvectors, which is guaranteed for symmetric matrices.