[Scilab-users] matrix of combinations

[scilab.20.browseruk at xoxy.net]

What are you going to do with this data once you've produced it?

If you are not going to use it straight away, then you would have to store it to disk, in which case there is no point in filling memory prior to writing it all out; just do one row at a time.

Conversely, if you are going to use it in the same run, do you actually need it all at any given time?

In many (even most) cases, depending how you are using the data, it is better to produce some subset -- say one row or other subdivision -- then use it, before discarding and looping back to produce the next subset.

If your application for this data is, or can be made to be, conducive to being subsetted this way, the effects on runtime can be dramatic. 
(Once recent application fell from 2 weeks to less than 12 hours.)

Just food for thought.

Buk

> -----Original Message-----
> From: scilab.browseruk.fa1c38e59e.paul.carrico#free.fr at ob.0sg.net
> Sent: Thu, 9 Jun 2016 18:47:58 +0200 (CEST)
> To: users at lists.scilab.org
> Subject: Re: [Scilab-users] matrix of combinations
exclusive)
> 
> Hi samuel,
> 
> (thanks)
> 
> I'm still under the latest stable Scilab release 5.5.x ... but it sounds
> a good opportunity to move to 6.0 one.
> 
> I've 2 servers (both under CentOS) having different features :
> - 2 CPU-s with 120 Go of memory
> - 8 CPU's with 30 Go of memory
> 
> the latest one is probably the more interesting for the current
> application :-)
> 
> I'll have a look to the scidoe module
> 
> Paul
> 
> ----- Mail original -----
> De: "Samuel Gougeon" <sgougeon at free.fr>
> À: "Users mailing list for Scilab" <users at lists.scilab.org>
> Envoyé: Jeudi 9 Juin 2016 18:10:36
> Objet: Re: [Scilab-users] matrix of combinations
> 
> 
> 
> Hello Paul,
> 
> Le 09/06/2016 17:28, Carrico, Paul a écrit :
> 
> 
> 
> 
> 
> Dear all
> 
> 
> 
> (not sure to really master the topic but …)
> 
> 
> 
> In my current study, I’ve 16 variables that I would like to perturbate 10
> times ; of course each variable is different from the others.
> 
> A1 = 10 + 2*rand(10,1,”uniform”)
> 
> A2 = 5 + 0.5*rand(10,1,”uniform”)
> 
> …
> 
> A16 = 0.2 +0.01**rand(10,1,”uniform”)
> 
> 
> 
> 
> 
> Except if I’m mistaken, I’ve about 16^10=1E12 combinations that is a
> (16xn) huge matrix M (n greater than 6E10) : how can I create such matrix
> ? It depends how you build a sample: is a sample a {A1(i), A2(j),.......,
> A16(x)} set?
> Then, there are 10^16 possibilities.
> You can build them with ndgrid(), provided that you have hexabytes of
> RAM, (really) many processors, and of course Scilab 6. Do you? :)
> Or you may have a look at DOE theory, and tools:
> https://atoms.scilab.org/toolboxes/scidoe
> 
> BR
> Samuel
> 
> 
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users

____________________________________________________________
FREE 3D EARTH SCREENSAVER - Watch the Earth right on your desktop!
Check it out at http://www.inbox.com/earth