[phenixbb] Random seeds?

Ralf W. Grosse-Kunstleve rwgk at cci.lbl.gov
Tue Aug 10 15:09:37 PDT 2010


> Okay, a fairly mundane question concerning random seeds in
> Phenix. Obviously they are random but any restrictions on how many
> digits phenix will accept?

It is determined by the random number generators we're using. Up to
9 digits should be fine. I'm not quite sure what the exact limit is.

> Any other restrictions?

Don't think so. Any integer up to a certain limit should work.

> Are they (it) only used for coordinate
> refinement

Any time the weight calculation is involved random numbers enter
through the assignment of random velocities for a short Cartesian
dynamics calculation. This is used to "shake" the model before
computing x-ray/neutron gradients and restraint gradients.
The weight is determined as the ratio of the norms of these gradients
(after outlier removal).

> or do the random seeds get used for other things as well?

The weight calculation and Cartesian or torsion-angle simulated
annealing are the main uses. Beyond that it takes a closer look
to know if random numbers enter into a particular procedure.

Ralf



More information about the phenixbb mailing list