Random seed

LAMMPS LAMMPS Mailing List Mirror
1

Hello all

I am using LAMMPS to run some DPD simulations. I want to run several independent runs with different random seeds. My simulation involve several thousand steps of equilibration (where I dont save detailed data) followed by production run. To make several independent runs would it be sufficient to just run the production runs with a different seed( this will save a lot of time)? Or I should run from the very beginning?

Thanks
Santo

2

Hello all

I am using LAMMPS to run some DPD simulations. I want to run several
independent runs with different random seeds. My simulation involve several
thousand steps of equilibration (where I dont save detailed data) followed
by production run. To make several independent runs would it be sufficient
to just run the production runs with a different seed( this will save a lot
of time)? Or I should run from the very beginning?

when starting different runs from the same initial configuration with
different random seeds, you will indeed save having to do the
equilibration, however, you cannot use the trajectories as completely
independent data, you still have to discard the first part until the
different trajectories are sufficiently diverged to consider them
uncorrelated.

whether this will save time versus starting completely independent
runs depends on the time it takes to equilibrate versus the time it
takes to decorrelate.

axel.

3

Thanks Axel, got your point.

Santo