How to reduce the executation time for the below operation

Question

SatyaPrakash Gupta 2020-4-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/519245-how-to-reduce-the-executation-time-for-the-below-operation

评论： SatyaPrakash Gupta 2020-4-22

Hi,

Loc = [m*n] ;  % size of X has 1500000*2 
image_spots = [m*n] ; % size of X has 500000*2
distance_thresh = 0.5;
mask = false([length(Loc),1]);
    tic,for i = 1:length(Loc)
        if distance_thresh > min(pdist2(Loc(i,:),image_spots))
            mask(i)= true;
        end
    end,toc;

the above for loop take atleast 12mins, is there a method or a way where i can reduce the times for the executation ?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

darova 2020-4-20

Improve your chances on answer

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2020-4-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/519245-how-to-reduce-the-executation-time-for-the-below-operation#answer_427064

I suggest that you redesign to use knnsearch https://www.mathworks.com/help/stats/knnsearch.html asking for 1 neighbour, and comparing that one resulting distance to the threshold.

knnsearch with euclidean distance function and as few as 2 dimensions, will construct a quadtree to be able to find the distances much faster. It becomes like a 2D binary search.

2 个评论
显示无隐藏无

SatyaPrakash Gupta 2020-4-20

actually it is bit complicated because i used a grid to extract and filter from 1500000 and after that i got 500000 data and now i have to accociated this data into actual location , therefore redesigning would be not helpful i guess.

it means , there is no other mechanism to reduce the executation code for the above code provided right ?

SatyaPrakash Gupta 2020-4-22