[Scilab-users] vectoriztion and sets of data

[Adrien Vogt-Schilb]

On 22/11/2012 22:59, Paul Carrico wrote:
>
> Thanks for your suggestion ; it's is interesting for another specific 
> development... unfortunately I don't think I cannot adapt that code 
> for my project !
>
> Indeed I need to work on sets of data (n rows of data at the same time 
> to select, to rearrange, to transform, to mean, etc. ...) on matrix on 
> size on X*n rows (X may be 100 000 ... ) ; currently I use loops as 
> described above ...
>
> ... the issue is to optimize the following line through vectorized "i" index:Search  =  max(abs(C([i.  *  8  -  7  :i.  *  8],:)))

Hi

Right.
Well a trivial improvment is still :

D=zeros((nr/8),1);

fori=1:(nr/8)

   D=max(abs(C((8*(i-1)+1):(8*i),:)))>0;  //oh, that's what you meant 
with "ok with the if"

end

But you want to get rid of the loop I gess.


Do you actually build the C matrix yourself as in the example?

I believe your example is equivalent to:

A=rand(8,6);

B=zeros(8,6);

C=[A;A;B;A;B;B;A];

D=[max(abs(A))>0; max(abs(A))>0; max(abs(B))>0; max(abs(A))>0; 
max(abs(B))>0; max(abs(B))>0; max(abs(A))>0]

or better to

isA = max(abs(A))>0;
isB =max(abs(B))>0;


C=[isA; isA; isB; isA; isB; isB; isA]

in other words: the idea is to write D while you build C, not once C is 
built entirely. Does that make sense for your actual code?


> Si let 's rack one's brains ...
>
> Thanks
>
> Paul
>
> *De :*users-bounces at lists.scilab.org 
> [mailto:users-bounces at lists.scilab.org] *De la part de* Adrien Vogt-Schilb
> *Envoyé :* jeudi 22 novembre 2012 16:12
> *À :* International users mailing list for Scilab.
> *Objet :*Re: [Scilab-users] vectoriztion and sets of data
>
> Hi again
>
> I am still not sure, but look at the output of
>
> A=rand(8,6)
>
> B=zeros(8,6)
>
> D = A > B
>
> D = bool2s (A > B)
>
> This might give you insight on how to solve your problem
> tell me if not
>
> On 22/11/2012 16:06, Paul Carrico wrote:
>
>     The original code looked like ... I'm trying to optimized it (ok
>     for the "if", but how to combine a set of data and vectorization
>     (??? Many trials but I failed)
>
>     /// initial matrix/
>
>     A=rand(8,6);
>
>     B=zeros(8,6);
>
>     C=[A;A;B;A;B;B;A];
>
>     [nr,nc]=size(C);
>
>     /// original code looks like/
>
>     D=zeros((nr/8),1);
>
>     fori=1:(nr/8)
>
>     search=max(abs(C((8*(i-1)+1):(8*i),:)));
>
>     if(search>0)then
>
>     D(i,1)=1;
>
>     else
>
>     D(i,1)=0;
>
>     end
>
>     end
>
>     ###################################################
>
>     mode(0)
>
>       
>
>     /// initial matrix/
>
>     A  =  rand(8,6);
>
>     B  =  zeros(8,6);
>
>     C  =  [A  ;  A;  B;  A;  B;  B;  A];
>
>     [nr,nc]  =  size(C);
>
>       
>
>     /// each set of 8 rows is analysed/
>
>     Search  =  zeros((nr/8),1);
>
>     ///i = ones((nr/8),1).* 1;/
>
>     i  =  [1:(nr/8)]';
>
>     /// Search = max(abs(C((8*(i - 1) + 1):(8 * i),:)));/
>
>     Search  =  max(abs(C([i.  *  8  -  7  :i.  *  8],:)))
>
>     ;
>
>     B  =  (find(Search  ==  0))';
>
>       
>
>     D  =  zeros((nr/8),1);
>
>     D(B,2)  =  1;
>
>
>
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20121123/b3eedcd7/attachment.htm>