Pseudoinverse

A+=(ATA)1ATA^+ = (A^T A)^{-1} A^T

Here, AA is a matrix with linearly independent columns and A+A^+ is its pseudoinverse. If AA is of size m×nm \times n, A+A^+ is of size n×mn \times m.

The idea is that if AA does not have an inverse, the pseudoinverse is the closest approximation. See also the least squares solution.

Last updated