stephenbrooks.orgForumMuon1Generalv4.33
Username: Password:
Search site:
Subscribe to thread via RSS
Stephen Brooks
2003-12-10 04:30:04
Not 4.32h?  No, well I actually added some features to the client this time instead of just bugfixing.  The RAR file is here and the ZIP file is here.  Barring any critical problems people have with this, I'll publish this on the main page in a day or two.

New features:
  • Running muon1.scr installs the graphical version as a screensaver

  • Optional periodic download (default= 3 days) of sample100 file, which is merged with results.dat

  • Missing CRLFs on results.* automatically fixed

  • Optimisation step type stored as #gen=x; in results, might be useful for analysis

...and several other small fixes/tweaks, for instance pressing both buttons on the mouse stops the panning and lets you mouseover particles to see their individual energies (this might develop into a bit more of a sensible GUI later).

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
excaliber[Free-DC.org]
2003-12-10 11:07:11
Do you recommend we download the sample file?  Im not using the best100 file (as discussed in the other thread), would it still be good to not use the sample100 file?

Thanks
Herb[Romulus2]
2003-12-10 11:12:14
I always merge the sample files with my own best of file, would this be circumvented then?

-------------------------------
I'd say more, but I can't reach the keyboard from the floor.
Stephen Brooks
2003-12-11 01:41:31
The program periodically merges sample100.txt (there is no best100.txt generated by me any more) with results.dat.  If that's what you were doing manually before, then you can set auto-merge to a number of days you like and it will do it.  If you were using some other method to add results to your results.dat, then it might be better to switch auto-merge off and continue doing that.

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
excaliber[Free-DC.org]
2003-12-11 03:26:32
But would you recommend us doing that?  Wont that lead to even more drastic inbreeding?  Or would it scientifically help?

Thanks
Stephen Brooks
2003-12-11 07:23:05
I want some people to frequently synchronise with the core results, and some people to be independent.  Having both is a way to prevent inbreeding and at the same time take advantage of the distributed nature of the project, since if a separate optimiser finds a better result, this becomes inserted into the sample results that people staying with the 'core' of the optimisation will use.

So in the end, it's up to you - flip a coin?

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
Lex_Luthor
2003-12-11 10:56:09
So for those of us that run multiple instances of the program on different machines- should we evenly distribute between them independently developed dats and those that will be continually updated?

"We're sorry to bother you at such a time like this, Mrs.  Twice.  We would have come earlier, but your husband wasn't dead then."
excaliber[Free-DC.org]
2003-12-11 11:39:33
Two questions, relating to this topic.

Firstly, say I want to work on my own 'breed'. Is there a way to synch my own computers to use the same results.dat file?

On the same idea, it would be a cool feature in later clients to specify a team breed.  Each team uses their own results.dat, so each team has their own breed.  This would be optional of course, but could lead to some interesting results (esp.  between the top several teams that have the most manpower.)
Stephen Brooks
2003-12-12 04:20:04
I've left the business of creating merged files for private groups of computers up to the users.  Some of the German teams I think do this regularly themselves, and probably have scripts you can borrow to do this.

The simplest level is to take all the results.dat files, append them into one big long file and replace the individual computers' results.dat by that.

HB Pencils, also sold as "Moron's Choice" Graphite Cigars.
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io RSS feed

Site has had 25100042 accesses.