rpi_logo
Differential Evolution?
Differential Evolution?
log in

Advanced search

Message boards : MilkyWay@home Science : Differential Evolution?

Author Message
Emanuel
Send message
Joined: 18 Nov 07
Posts: 280
Credit: 2,442,757
RAC: 0

Message 30400 - Posted: 9 Sep 2009, 7:21:32 UTC

On August the 28th, Travis told us he was starting up a new type of searches - and that he would explain them in more detail in a forum post. I'm putting this up as a bit of a reminder - could you give us more information about the new searches?

sygopet
Send message
Joined: 18 Apr 09
Posts: 5
Credit: 3,250,036
RAC: 4,798

Message 31136 - Posted: 18 Sep 2009, 12:22:15 UTC - in response to Message 30400.
Last modified: 18 Sep 2009, 12:25:23 UTC

On August the 28th, Travis told us he was starting up a new type of searches - and that he would explain them in more detail in a forum post. I'm putting this up as a bit of a reminder - could you give us more information about the new searches?

. . and what is the significance of units beginning gs_?
also, while I accept that computer users sometimes have an excessive desire to make up new words to describe their trade, what on earth does "constrainted" indicate?

Emanuel
Send message
Joined: 18 Nov 07
Posts: 280
Credit: 2,442,757
RAC: 0

Message 31147 - Posted: 18 Sep 2009, 16:46:43 UTC - in response to Message 31136.
Last modified: 18 Sep 2009, 16:55:01 UTC

. . and what is the significance of units beginning gs_?
also, while I accept that computer users sometimes have an excessive desire to make up new words to describe their trade, what on earth does "constrainted" indicate?

Hehe, I never noticed that before.. I wonder if it's a typo? And Genetic Search I am familiar with - basically parameter files from previous iterations are combined and 'transcribed' with some combination of 'mutations', much like the way genetics works in nature. Since this is a random process, it's up to the server to decide how far to let parameters deviate from the current best before cutting off the dead ends. The less room for seemingly useless mutations, the faster it'll converge - but you risk missing out on the most radical improvements.

Differential Evolution seems very interesting though - although it appears to take more time to converge than the Particle Swarm, it's getting some even better results. (the same can't be said for the Genetic Search at the moment)

Emanuel
Send message
Joined: 18 Nov 07
Posts: 280
Credit: 2,442,757
RAC: 0

Message 31351 - Posted: 22 Sep 2009, 17:33:59 UTC

So all the new WUs are using Differential Evolution.. Can we get some information on these new WUs? Travis? John?


Post to thread

Message boards : MilkyWay@home Science : Differential Evolution?


Main page · Your account · Message boards


Copyright © 2019 AstroInformatics Group