[Gptuser] Tricks regarding optimization

[Fay Hannon]

Info from Pulsar regarding a question on what can be changed during an
optimization run because I wanted to increase particle number as it
increases generation number:

 

"

Regarding .in, .mr and .mgo files:

- You can change the .in file during an mgo optimisation: New evaluations
will just use the new .in file. I sometimes do that for the number of
particles where I start for example at nps=1000 in the inputfile, and
simultaneously I set an mgo constraint nps.minimum = 10000. After a while I
increase particle number to say 2000, few hours later to 5000 and eventually
to 10000. The constraint is required to make sure that super-good results
obtained with less particles are eventually all kicked out. 

- Changing the .in file of a single GPT run has no effect. The inputfile is
read once, done.

- Same for mr file.

- Please note that if you have many GPT runs with MR, changing the .in file
will result in some runs using the old, and some runs using the new
inputfile. Typically this is not the smartest thing to do.

"

 

Fay

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/gptuser/attachments/20180308/354a1bdb/attachment.html>