Using a vector of indices to add values to a matrix

조회 수: 2 (최근 30일)
Christopher
Christopher 2014년 3월 2일
답변: Jos (10584) 2014년 3월 5일
I am trying to vectorize a piece of code which has this form:
for m = 1:1:numel(N)
y = N(m,1);
x = N(m,2);
val = N(m,3);
P(y,x) = P(y,x)+val;
end
So, there is the matrix of column vectors N which hold coordinate information and a value. Using the coordinates the corresponding values are added to the matrix P.
How can this be done efficiently? Thanks.

채택된 답변

Paul
Paul 2014년 3월 2일
편집: Paul 2014년 3월 2일
xs = N(:,1);
ys = N(:,2);
val = N(:,3);
siz = size(P);
ind = sub2ind(siz,ys,xs);
P(ind) = P(ind)+val.';
  댓글 수: 2
Jos (10584)
Jos (10584) 2014년 3월 2일
Note that the addition will not work properly for duplicates of (xs,ys) / ind
Christopher
Christopher 2014년 3월 5일
편집: Christopher 2014년 3월 5일
I think I have found that you are right. The result for the loop and the vectorized operation are completely different. Perhaps because there is a lot of repeated indices (thus addition) in my code. Does anyone know of a workaround???

댓글을 달려면 로그인하십시오.

추가 답변 (1개)

Jos (10584)
Jos (10584) 2014년 3월 5일
ACCUMARRAY is, by far, the easiest approach:
N = [1 2 7 ; 1 2 4 ; 2 3 5 ; 2 1 6]
P = accumarray(N(:,1:2),N(:,3))

카테고리

Help CenterFile Exchange에서 Resizing and Reshaping Matrices에 대해 자세히 알아보기

제품

Community Treasure Hunt

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

Start Hunting!

Translated by