Recall our equation from least-squares:
Let’s use the QR factorization: .
Because is upper triangular we have
Thus, we get the equivalent linear system
Let’s solve for :
using .
The final solution is
Recall the normal equation:
The condition number is reduced from
to
This is a huge improvement!
Note that this method requires to be non-singular. This is equivalent to saying that should be full column rank.
The computational cost is .
Least-squares problems, QR factorization, QR using Householder transformations, QR using Givens transformations, Method of normal equation