What Is the Moore-Penrose Pseudoinverse?
The Moore-Penrose pseudoinverse, denoted A⁺, is a generalization of the matrix inverse. While a standard inverse exists only for square, non-singular matrices, the pseudoinverse exists for any matrix regardless of shape or rank. It is the unique matrix that satisfies the four Moore-Penrose conditions.
How Is the Pseudoinverse Calculated?
For an m x n matrix A, the pseudoinverse depends on the relationship between m and n:
Overdetermined (m > n)
More rows than columns. The left pseudoinverse is used.
A⁺ = (AᵀA)⁻¹Aᵀ
Underdetermined (m < n)
More columns than rows. The right pseudoinverse is used.
A⁺ = Aᵀ(AAᵀ)⁻¹
Square Matrix (m = n)
If invertible, the pseudoinverse equals the standard inverse.
A⁺ = A⁻¹
Moore-Penrose Conditions
The pseudoinverse A⁺ is the unique matrix satisfying all four conditions:
- AA⁺A = A -- A⁺ acts as a weak inverse.
- A⁺AA⁺ = A⁺ -- A acts as a weak inverse of A⁺.
- (AA⁺)* = AA⁺ -- AA⁺ is Hermitian (self-adjoint).
- (A⁺A)* = A⁺A -- A⁺A is Hermitian (self-adjoint).
Applications of the Pseudoinverse
- Least Squares: Solving overdetermined linear systems Ax = b with x = A⁺b.
- Data Fitting: Linear regression and curve fitting in statistics.
- Robotics: Computing inverse kinematics for redundant manipulators.
- Signal Processing: Beamforming and channel estimation in communications.
- Machine Learning: Ridge regression, neural network training, and PCA.
Tips for Accurate Computation
- For numerical stability, SVD-based methods are preferred over the direct formula.
- Check the condition number of your matrix to assess numerical sensitivity.
- The pseudoinverse of a zero matrix is the zero matrix (transposed).
- For rank-deficient matrices, the direct formulas may fail -- use SVD instead.