How to speed up this code? Or vectorize?

4 次查看(过去 30 天)
hi, i got the following code. Is there a way to completly vectorize it ? or rather a better way of programming this for faster execution in e.g fmincon.
c=positive_int;
% create alpha for case n=0;
j=0:50;
alpha(0+1,j+1)= gamma((c.*j)+1)./gamma(j+1);
cache=zeros(51,1);
for n=1:1:10
for j=n:1:50
% make array of new_alpha(j) in a matrix with old_alpha in column n
% with increasing index the the new_alpha, new_alpha is a sum of old_alpha from old column with index * function(value_index)
% e.g new_alpha(1)=old_alpha(1)*function(with value1);
% new_alpha(2)=old_alpha(1)*function(with value1)+ old_alpha(2)*function(with value2); ....
% and that for every n
for m=n-1:1:j-1
% recursive using alpha:
cache(m+1)= alpha(n-1+1,m+1)*gamma((c*j)-(c*m)+1)/gamma(j-m+1);
end
alpha(n+1,j+1)=sum(cache);
cache(:)=0;
end
end
So far it got this.
c=positive_int;
j=0:50;
alpha(j+1,0+1)= gamma((c.*j)+1)./gamma(j+1);
variable=50;
[mValues, nValues] = meshgrid(0:variable, 0:variable);
mask = nValues <= mValues;
for n=1:1:10
% Operation for one new alpha_array:
[j,m]=meshgrid(n:1:50 ,n-1:1:49);
cache= alpha(m(1:50-n+1)+1,n-1+1).*gamma((c.*j)-(c.*m)+1)./gamma(j-m+1);
mask(51-n,:)=[];
mask(:,51-n)=[];
cache=cache.*mask;
cache(isnan(cache))=0;
cache=sum(cache);
alpha(n+1:end,n+1)=cache;
end
But that seems a bit complex coded and is not that huge improvement.
  1 个评论
Jan
Jan 2019-2-1
编辑:Jan 2019-2-1
This let your array shrink iteratively, what is a don't for efficient programming:
mask(51-n,:)=[];
mask(:,51-n)=[];
This happens 10 times only, so it will not matter much.
Please do not let the readers guess, what "positive_int" is. Post running code.

请先登录,再进行评论。

回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Matrix Indexing 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by