Pseudoinverse Calculator

Calculate Moore-Penrose pseudoinverse for any matrix with step-by-step solutions

Matrix Input

Select matrix dimensions. Currently optimized for 2×2, 2×3, and 3×2 matrices.

[
]

Enter matrix elements. Use decimal numbers for fractional values.

Pseudoinverse Results

Original Matrix A:

[1.00002.0000]
[3.00004.0000]

Pseudoinverse A⁺:

[-2.00001.0000]
[1.5000-0.5000]

Method Used:

Regular inverse (matrix is invertible)

Note: Since the matrix is invertible, the pseudoinverse equals the regular inverse.

Properties Verified:

  • • A⁺ has dimensions 2×2
  • • The pseudoinverse is a generalization of the matrix inverse
  • • Satisfies the Moore-Penrose conditions
  • • For invertible matrices: A⁺ = A⁻¹

Example: 2×3 Matrix

Try this example:

Matrix A = [[1, 3, 2], [4, 3, 3]]

This matrix has linearly independent columns, so we can use A⁺ = (AᵀA)⁻¹Aᵀ

Calculation Methods

1

Left Pseudoinverse

A⁺ = (AᵀA)⁻¹Aᵀ

For linearly independent columns

2

Right Pseudoinverse

A⁺ = Aᵀ(AAᵀ)⁻¹

For linearly independent rows

3

SVD Method

A⁺ = VS⁺Uᵀ

General case using SVD

Properties

Generalizes the concept of matrix inverse

Exists for any matrix (including non-square)

A⁺ has transposed dimensions of A

If A is invertible, then A⁺ = A⁻¹

Used for solving overdetermined systems

Understanding the Moore-Penrose Pseudoinverse

What is the Pseudoinverse?

The Moore-Penrose pseudoinverse A⁺ is a generalization of the matrix inverse for matrices that may not be invertible. It exists for any matrix and provides the "best" solution to linear systems that may have no solution or infinitely many solutions.

Key Properties

  • Universal existence: Every matrix has a pseudoinverse
  • Dimension property: (m×n) matrix has (n×m) pseudoinverse
  • Reduces to inverse: A⁺ = A⁻¹ when A is invertible
  • Least squares: Minimizes ||Ax - b||² in overdetermined systems

Calculation Methods

Case 1: Linearly independent columns

A⁺ = (AᵀA)⁻¹Aᵀ

Case 2: Linearly independent rows

A⁺ = Aᵀ(AAᵀ)⁻¹

General case (SVD)

A = USVᵀ → A⁺ = VS⁺Uᵀ

Applications

  • Linear regression: Finding best-fit lines and curves
  • Data fitting: Approximating solutions to inconsistent systems
  • Image processing: Reconstruction and denoising
  • Machine learning: Principal component analysis
  • Control theory: System identification and optimization