...invertible

The reader familiar with the statistical literature will recognize in C the model covariance matrix. The statistical roots of the least-square optimization are in the method of maximum likelihood. Connecting this statistical estimation method with least squares requires an assumption of uncorrelated additive noise with zero-mean Gaussian distribution.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...Claerbout (1994)

The Fortran-90 version of the electronic book

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...imaging

available at http://sepwww.stanford.edu/sep/biondo/GP291/Notes/Ps/notes-latest.ps

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...Fortran-90

Sidestepping the religious C++ versus Fortran war, I would like to point out that an object-oriented design is more important than the actual implementation. The design should be easily transferrable from Fortran-90 to other advanced languages.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...method

Our current library has about 6 other functions of the same interface, implementing different linear and nonlinear optimization methods.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.