Question regarding GMP/MP implementation

[maxanier]

Hi,
first of all thank you for sharing this Python implementation of MP/GMP DPD. I was successfully able to apply this to my PA in the lab.

When studying the code, I was stumbling across the 0.000001*np.eye(X.shape[1]) in the coefficient extraction.

pyrfdpd/pyrfdpd/volterra/gmp.py

Line 73 in dc22697

coef = np.linalg.pinv(XH.dot(X) + 0.000001*np.eye(X.shape[1])).dot(XH).dot(y_target)

I assume this line implements equation (29) from the paper by Morgan et al. (10.1109/TSP.2006.879264):
$w=(Y^H Y)^{-1} Y^H x$, where w are the coefficients, $Y$ equals X from the code and $x$ equals y_target.
However, this additional part helping with "stability" of the code or is there another reason?

And is it correct, that this approach of coefficient extraction is called "least-square".

Many Thanks