Pseudoinverse Geometry Lab

Too many equations, too many unknowns, and what A⁺ actually chooses. The pseudoinverse finds the closest reachable output, then chooses the simplest input that reaches it.

Closest valid output. Smallest valid input.

$The Moore-Penrose pseudoinverse A^{+} b is the minimum-norm least-squares solution to A x = b . When the system is too tall, it projects b onto what A can reach. When it is too wide, it picks the shortest x among infinitely many. The SVD shows why this is the safe universal definition: A^{+} = V Σ^{+} U^{⊤} .$

$Too tall: project b onto Col (A)$

$A maps one input number into a line in output space. Most targets b are not on that line.$

A \in R^{2 \times 1}

TheoremPath route

This lab sits between Matrix Mechanics and SVD. Matrices are maps, projection is closest-point geometry, and SVD is the safe universal way to compute the pseudoinverse.

Vectors, Matrices, Linear Maps Inner Products & Orthogonality Matrix Operations SVD Linear Regression PCA

TheoryMatrix Operations and PropertiesThe reference page for the Moore-Penrose definition and least-squares formulas TheorySingular Value DecompositionWhy A⁺ = VΣ⁺Uᵀ and what condition number actually measures Prereq LabMatrix Mechanics LabDeterminant, rank, row reduction, eigenvectors, and the SVD intuition this lab depends on

TheoremPath · 641 topics · Interactive demos

Pseudoinverse Geometry Lab

Too tall: project b onto Col(A)

TheoremPath route

$Too tall: project b onto Col (A)$