Pages: [1]
  Print  
Author Topic: Best PC configuration for Rapid-I?  (Read 1818 times)
Liliput
Newbie
*
Posts: 5


« on: January 03, 2009, 12:16:10 PM »

Hi,

can we please discuss, which will be the best configuration to buy to speed up Rapid-I work with large example sets?

I recently use a Pentium4 - 3.0Ghz PC with 1,5Gb RAM and fast SATA HDD on WinXP.
My task is to extract rules via the RuleLearner for a large dataset (>65000 examples) :-( I do not want wight the set or reduce the # of the examples and had run a test, which i waited for 24h and i founded still working.

So can anyone please explain: Will a Core QUAD CPU with 4GB RAM do it much faster so that it will be worht to buy it? Is Rapid ready to use those HT multitasking skills or it will be better to buy 8GB RAN for the old configuaration?

Thanks and happy new year Smiley
Logged
haddock
Hero Member
*****
Posts: 853



WWW
« Reply #1 on: January 03, 2009, 04:43:56 PM »

Quote
So can anyone please explain: Will a Core QUAD CPU with 4GB RAM do it much faster so that it will be worht to buy it? Is Rapid ready to use those HT multitasking skills or it will be better to buy 8GB RAN for the old configuaration?

It is a popular misconception that things have an absolute value, and a silly one. A thing is worth what someone will pay for it, not one cent more. Banks thought that subprime loans were worth what their models said, oops. All I can say is that more memory helps, I work on 12G, and muti-core helps on optimisations and validations,  but only for the commercial version of RM. So you might, as they say, be missing the elephant in the room, in that the commercial versions cost more than the hardware necessary to run them.
Logged

Where is the wisdom we have lost in knowledge?
Where is the knowledge we have lost in information?

T.S.Eliot ~ Choruses from the Rock 1934
Sebastian Land
Administrator
Hero Member
*****
Posts: 2426


« Reply #2 on: January 05, 2009, 12:31:18 PM »

Hi,
as haddoc already said: Multiple threads cannot be automatically supported since learning tasks are highly interdependet on previous results. But there are some learner as the rule learner providing some parallelizable parts. But this is only included in the enterprise edition.
So if you want to have a fast computer for community edition buy a high performance dual core with fast ram. It cannot be wrong to buy lots of RAM, but if the problems fits into your 1.5 GB, 2 GB should do the work, too. If you want to use more than 3 GBs or RAM you have to use a 64bit operating system.

I personally would rather tune my process setup, since a new (single thread) computer will not speed up the computation more than twice. While changing the setup might do the trick. Perhabs the rule learner is somehow caught in the very noisy remaining examples or overfits misserably. The resulting rules then will not tell you anything and hence the generation might be stopped before.

If you don't need the rules as human readable model, you might change the learning algorithm.

Another important fact on the rule learner is, that the number of attributes has heavy impact on the runtime.


Greetings,
  Sebastian
Logged
Liliput
Newbie
*
Posts: 5


« Reply #3 on: January 12, 2009, 12:15:25 PM »

Thank you for the answers! It helps a lot to know such details. Like they say the devil hides in the details ....
I will improve the RAM speed and size :-)
Logged
Pages: [1]
  Print  
 
Jump to: