Can I provide a starting point y to one of the eigen's solver for Ax = b?
I want to get an exact solution x that is close to the starting point y.
When would solving Ax = b with Newton's method (without line search) not converge?
Thanks.
Solve Ax = b for a solution that minimize L2 norm of x - y.
Given a y, that means minimizing L2 norm of x - y, subjected to the constraint Ax = b.
By Lagrange multiplier I get the following block matrix equation.
The vector u are one half of the lagrange multipliers.
/ A 0 \ / x \ / b \
| | | | = | |
\ I A^T / \ u / \ y /
Probably not the fastest way.