RE: random seed

From: JC Gumbart (gumbart_at_ks.uiuc.edu)
Date: Tue Sep 27 2005 - 20:30:05 CDT

The number chosen by NAMD as the seed value is just the Unix time (measured
in seconds from 1970, I believe). This is why they are all so close.

This is a very common method for generating random numbers. I suspect there
should be no long time correlation between pseudo-random numbers produced by
seeds that are close as compared to ones that are far apart; otherwise this
would be a very poor method, obviously.

Looking at the formula used by NAMD would certainly confirm this.

-----Original Message-----
From: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] On Behalf
Of ioana_at_pegasus.arc.nasa.gov
Sent: Tuesday, September 27, 2005 8:14 PM
To: namd-l_at_ks.uiuc.edu
Subject: namd-l: random seed

Hi,

I have a question about the random seed generated by NAMD.
By definition if it is defined in the configuration file it can be any
positive integer. However if it's not defined in the config file, the
program writes it out in the output file but there it's always a number in
the order of 1125770000 or so. Why is that? Why am I not seeing smaller
numbers (say 100 or 2100?).

Also how different can I consider that there are runs (given the same
initial conditions, only random seed is different) whith the following
initial seeds:

./r1/r1.o3926:Info: RANDOM NUMBER SEED 1127443123
./r2/r2.o3927:Info: RANDOM NUMBER SEED 1127443123
./r3/r3.o3928:Info: RANDOM NUMBER SEED 1127443127
./r4/r4.o3929:Info: RANDOM NUMBER SEED 1127443131
./r5/r5.o3930:Info: RANDOM NUMBER SEED 1127443133

Are they reliable for statistical purposes?

Thanks,
Ioana

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:10 CST