x′ is measured location and M is unknown matrix
x′=MxLeast Square Error Function f(xi;p) is predicted location
ELS=Σi∣∣f(xi;p)−xi′∣∣2p is parameter
p^=argminpELS
General form of linear least squares in matrix form
ELS=∣∣Ax−b∣∣2=xT(ATA)x−2xT(ATb)+∣∣b∣∣2When you solve derivative
(ATA)x=ATbx=(ATA)−1ATbbut inverse of Matrix sucks