| Re: Randomize array order [message #55019 is a reply to message #55017] |
Thu, 26 July 2007 06:40   |
Conor
Messages: 138
Registered: February 2007
|
Senior Member |
|
|
On Jul 26, 9:30 am, Allan Whiteford
<allan.rem...@phys.remove.strath.ac.remove.uk> wrote:
> Conor wrote:
>> Hi everyone!
>
>> Anyone know an efficient way to randomize an array (I have a
>> sorted array that I want unsorted). Initially, I tried something like
>> this:
>
>> array = findgen(1000000)
>> unsort = array[sort(randomu(seed,1000000))]
>
>> It works, but sorting on a million elements is rather slow. Anyone
>> know a faster way?
>
> Conor,
>
> Is it a million elements you want to do?
>
> The following scales better:
>
> pro shuffle,in
> b=long(n_elements(in)*randomu(seed,n_elements(in)))
> for i=0l,n_elements(in)-1 do begin
> tmp=in[i]
> in[i]=in[b[i]]
> in[b[i]]=tmp
> end
> end
>
> but on my machine, a million elements is around about where it starts to
> become as efficient as yours. For 10 million elements the above is a bit
> (17.05 seconds vs 12.92 seconds) but for 1 million elements they both
> come in at around 1.2 seconds (1.15 seconds vs 1.26 seconds). The above
> will scale as pretty much O(n) since it doesn't do any sorting but it
> takes a hit in the practical implementation because of the loop in
> IDL-space. Your suggestion will scale worse than O(n) but it seems the
> overlap in the two methods is exactly where you want to work.
>
> Maybe my loop can be made more efficient in practical terms but I don't
> think this is any better algorithm in terms of scaling (hard to imagine
> anything that could go faster than O(n) to randomise n things).
>
> Probably not helpful but I thought it was interesting that the
> cross-over is exactly where you want to work. But, maybe I should get
> out more if I think that's especially interesting.
>
> Thanks,
>
> Allan
Thanks for the suggestions guys! I'll have to play around and see
what works best.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
