Vectorizing and preallocating these "for" loops

Question

Johnathan 2014-4-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/124359-vectorizing-and-preallocating-these-for-loops

编辑： Azzi Abdelmalek 2014-4-3

I'm a fairly new MATLAB user and I'm having trouble with speeding up my code. There are two m-files that need to be improved for this assignment I am doing. My professor put emphasis on addressing the "for" loops in the code, but no matter what I tried I'm not having much success.

The first m-file follows:

N = length(x(:,1));    %# points
xScaled = [x(:,1)-min(x(:,1)) x(:,2)-min(x(:,2))];  %subtract minimum from each [x,y]
xGrid   = xScaled./h;                                 %normalize
xGrid   = floor(xGrid);                               %quantize to grid
xGrid   = xGrid+1;                                    %add 1 because matlab is 1's based
%the lowest grid value will be (1,1);
%this can be converted to a linear index
%determine number of grid cells in x and y directions
nMax      = max(xGrid);
xIndex    = xGrid(:,2)+nMax(2)*(xGrid(:,1)-1);  %index = yCoord+height*xCoord
influence = zeros(N,1);
for ind1 = 1:N
  tempInfluence = 0;
  neighboorhoodPoints = findNeighboorsv2(ind1,x,xGrid,xIndex);
    for ind2 = 1:length(neighboorhoodPoints(:,1)); %only loop over nearby particles
        tempInfluence = tempInfluence+computeInfluence(x(ind1,:),neighboorhoodPoints(ind2,:),h); 
        if (x(ind1,1)~=neighboorhoodPoints(ind2,1)) %make sure they are not the same point
        tempInfluence = tempInfluence+computeInfluence(x(ind1,:),neighboorhoodPoints(ind2,:),h);
        end
    end
   influence(ind1) = tempInfluence;
end

And the second m-file follows:

%find which bin it is in
targetBin = xGrid(pt,:);
nMax      = max(xGrid);
neighboorhoodPoints = [];
%gather points in adjacent bins
for ind1 = [targetBin(1)-1:1:targetBin(1)+1]
  for ind2 = [targetBin(2)-1:1:targetBin(2)+1]
        if (ind1>0 && ind1<=nMax(1) && ind2>0 && ind2<=nMax(2))  %only check over valid grid cells 
            %determine equivlant index
            index = ind2+nMax(2)*(ind1-1); %convert ind1 & ind2 into 
            neighboorhoodPoints = [neighboorhoodPoints; x(xIndex==index,:)];  %find returns the indices where the expression is true
        end
    end
end