Eigenvectors and Eigenvalues Visualized: Intuition, Computation, and Eigenbasis

Overview

The video presents an intuitive geometric view of eigenvectors and eigenvalues, explaining how certain vectors are stretched or squished when a linear transformation is applied, while others are rotated away from their span.

• eigenvectors stay on their span and scale by the eigenvalue
• eigenvalues measure the stretch or compression factor
• diagonal matrices arise when basis vectors are eigenvectors

How to Think About Eigenvectors and Eigenvalues

The video begins by framing a two dimensional linear transformation as a matrix whose columns describe where the standard basis vectors land. Most vectors do not stay on their original span after the transformation; they are rotated or skewed. However, special vectors, called eigenvectors, remain on their own span, and are simply scaled by a factor called the eigenvalue. In the illustrated 2D example, the x axis vectors are stretched by a factor of three and a diagonal line is stretched by a factor of two, while vectors not aligned with these spans generally rotate away from their original lines.

The Core Equation and How to Compute It

Symbolically, the eigenvector-eigenvalue relationship is A v = λ v, where A is the matrix for the transformation, V is an eigenvector, and λ is the eigenvalue. To analyze this, the right side can be written as a matrix times V, namely λ times the identity I times V. Subtracting this from both sides and factoring V leads to (A − λ I) V = 0. A nonzero eigenvector V exists precisely when the determinant det(A − λ I) = 0. This condition yields a polynomial in λ whose roots are the eigenvalues.

Worked Examples and Geometric Insights

Several classic examples are discussed. A two by two matrix with columns giving the action produces eigenvalues by solving det(A − λ I) = 0. In a rotation in the plane, there are no real eigenvectors because a pure rotation never leaves vectors on their spans. In contrast, a shear transformation fixes the x axis so every vector on that axis is an eigenvector with eigenvalue 1. A transformation that scales every vector by 2 has a single eigenvalue 2 with every vector in the plane being an eigenvector.

Eigenbasis and Diagonalization

When a basis consists entirely of eigenvectors, the transformation is represented by a diagonal matrix with eigenvalues on the diagonal. Diagonal matrices are much easier to power up, because applying the matrix repeatedly simply raises each eigenvalue to a corresponding power. If a matrix has a complete eigenbasis, you can perform a change of basis to diagonalize, compute powers there, and transform back to the original coordinates. Not all transformations admit a full eigenbasis, as in the shear example, but when such an eigenbasis exists it makes many matrix operations exceptionally pleasant.

Closing Reflections and Next Steps

The video closes by linking these ideas to higher dimensional intuition and previews an upcoming topic on abstract vector spaces, inviting the viewer to explore an eigenbasis puzzle that demonstrates surprising outcomes of diagonalization in action.