fractional power basis issue on restart

[Vladimir_Konjkov]

Hello CASINO developers

I've tried to use fractional power basis in the form r/(r+a) power in the Rank: [ 2, 0 ] term and in the Rank: [ 2, 1 ] for e-e distance only.
I successfully optimized JASTROW coefficients by EMIN, but when I try to restart calculation with optimized coefficients I've got unphysical E value.

example for B atom in attachment

Best Vladimir.

P.S.
same for 1/(r+a) power and r/(r^b+a) power in the Rank: [ 2, 1 ] for e-e distance. I tried to find a balance between the VMC optimization accuracy and the time of JASTROW evaluation in DMC calculation.
I wrote down a list of arbitrary combinations of JASTROW basic functions and cutoffs, then ran an automatic script to compute DMC timing for them all. Unfortunately, I have not even reached the middle of the list.

[Pablo_Lopez_Rios]

Hi Vladimir,

I see that in your first run you get the warning:

Code: Select all

Warning: [REDO_GAUSSIAN_ELIMINATION] non-linear parameters produce 1 zero(es)
 in the coefficient matrix for the constraint equations.  We recommend that
 you stop this calculation, change the values of any basis parameters involved
 to non-integer values, all slightly different from each other, and re-run to
 avoid possible problems.  This run will continue in the hope that the problem
 will go away during optimization.  The affected equations and parameters are
 listed below.

Zero #1 encountered at equation #17, parameter #26.

which hints at a problem with solving the constraint equations at the initial parameter values.

I would suggest you set the initial values of a and L to different, non-integer values, e.g., 3.1, 3.2, 3.3 and 3.4 (or start them closer to their optimized values if you want). This ought to work around the issue.

Best,
Pablo

[Vladimir_Konjkov]

Hello Pablo.

Oh magic! How did you know?

now I can compare the speed of the calculation, thanks.

Vladimir.