DPAD - best muon percentage question

[Zbox]

I'm a little confused as to why it seems mine is crawling up ever so slowly... (currently mine is pushing ~2.0 when the real highs are ~9.9)

Is this caused by using/not using the sample results?

 

[Confused]

If you use the sample results, then it will use the information from the sample results as a "guide" of where to start working, and will gradually make a better and better design.

If you don't use the sample results, then the client will do all this work itself. You may not be getting the higher muon percentages, but you are more likely to go down different paths, and make different designs in the process, which could or could not be better.

It's all going to the same thing in the end, which is the best design. How you get there is another matter! 🙂


Garry

 

[DanC]

It would appear you're not using the sample results. (you should)

Go to your config.txt file, and locate this line:
Download sample results file after a number of days (0=don't): 1

You can make it any number you want - but if it's "0" - you don't use them. I like "1" to stay current. 🙂

 

[Zbox]

I guess that's what I'm confused about then... each of my config.txt files read:
Download sample results file after a number of days (0=don't): 3

which should be using the samples, correct? but per my results, it seems like I'm not using them...

I read something on Stephen's board about the sample file being broken?

 

[JonB]

Zbox, I had noticed you had done LOTS of work units per Mpts. That explains it.
The randomness factor of the genome based client means that you could (by not using the sample results file) find some new and novel approach with a high percentage. Those of us using the sample result file will find incrementally better percentages, but probably not anything that makes you sit up and go WOW !!
The Mpts is what stats are based on. Your results.txt file probably reaches the 100k file size pretty quickly.

anyway, keep on crunching. Great to have you with us. I'm a relative newcomer to DPAD, but am dasm glad to be here.

 

[Zbox]

so would you say i'm not using the sample file despite my client being configured to do so? 😉

 

[DanC]

It would appear not. You however CAN download the 500-best, and save it as "config.dat" and it'll assist in better scores. Just don't name it "results.dat" - that's your current results file (as the name would imply) 😉

 

[Zbox]

will the client use config.dat in spite of a sampleresults.dat in the same directory?

does muon1.exe need to be restarted once config.dat is in place?

and would i need to change anything in config.txt?

I ask too many questions huh? 😉

 

[DanC]

It's actually config.txt - and they're not related... so yes - it does.

After making any changes... you do indeed need to stop, and restart.

Nothing need be changed in config.txt to use the sample results (best500) file.

 

[Zbox]

I haven't heard about a "config.dat" file until your post... was that a typo or am I just really confused?

I'm assuming I would replace "sampleresults.dat" with the 500k sample file from the website...

and would I be able to change this line in "config.txt" to allow future use of the sample file rather than downloading the file manually and renaming it as a one-time fix?

Sample file URL: http://stephenbrooks.org/muon1/sample100.txt
to:
Sample file URL: http://stephenbrooks.org/muon1/sample500.txt

But right now I'm convinced the sample files aren't being utilized properly because of missing checksum values. See this thread:
sample files broken

Is it possible others aren't noticing this problem (if it does indeed exist) because their clients already have successful high muon percentage in their "results.dat" to work from?

Thanks for all your help, DanC... I apologize for so many questions; just trying to familiarize myself with the project right now...

-z

 

[amdxborg]

zbox, i think the broken files maybe has something to do with it. I can mail you the results.dat and sampleresults.dat I've got, If you like. 😉

 

[Overkiller]

guys, you also have to remember a few other facts about results.dat & sample results & mpts


The higher the transfer value the higher the Mpts genereated. So a 6.5% transfer will be > a 2.3% transfer which is < than a 9.9% transfer.

also, the sample results is not simply just the top 100 transfers b/c that would be only through one strain and be counter-productive to the goals. The sample results file actually has various *leads* & strains along with a few lower end transfers to add to the variability...

oh, and if u check ur results.txt as results come in you'll see that the program still does run completely 100% random trials quite often (that is why a person can see a 9.9 % transfer followed by 6 0.5% transfers)

🙂

 

[Zbox]

apparently the sample files that the config.txt was pointing to by default are moved/renamed on the webserver:

these:
http://stephenbrooks.org/muon1/sample100.txt
http://stephenbrooks.org/muon1/sample500.txt

ought to be:
http://www.stephenbrooks.org/muon1/samplefiles/solenoidsto15cm_100.txt
http://www.stephenbrooks.org/muon1/samplefiles/solenoidsto15cm_500.txt

still they are missing checksum values, so I'm not too concerned about implementing them at this point. woo, i'm dominating the low end of the 500k sample file at least ;P

 

[DanC]

Sorry - my bad. I meant config.txt. 😱

edit:
I just realized everything I wrote was pretty much crap - (after reading earlier posts) - so I whacked it. 😉

 

[JonB]

so, what does it mean when the DPAD command line client says:

Quarantined result has now been run 2 of 5 times
New simulation
Rechecking quarantined result

the prior two runs were both over 304 Mpts.

 

[DanC]

Jon -
That's a GOOD thing. It means you've found a new "best-so-far" result, and it's re-checking the simulation to verify.

 

[Overkiller]

and the more times it re-checks the more MPts it generates :beer: