# an Alternative funtion which is faster than "ismember"

10 views (last 30 days)
Ahmet Hakan UYANIK on 24 Nov 2022
Commented: Bruno Luong on 24 Nov 2022
Hello everybody,
I was using ismembertol with XY(Nx2) and xy(Mx2). However code never ends due to the enormous amount of data(N=400million M=80mil.).
Is there any way that I can speed this function. (The matrices are not unique)
[LIA,~]= ismembertol(XY,xy,0.00001,'ByRows',true,'OutputAllIndices',true);
Bruno Luong on 24 Nov 2022
I post a short code as answer so you can accept it. Thanks

Bruno Luong on 24 Nov 2022
If XY is gridded coordinates, then you can use discretize or simple division if they are uniform to determine which grid the river point belong to.
% Generate some toy fake data
xgrid = cumsum(randi(5,1,10))
x = min(xgrid)+rand(1,10)*(max(xgrid)-min(xgrid))
midpoints = (xgrid(1:end-1)+xgrid(2:end))/2;
x_edges = [-Inf midpoints Inf];
iclosest_x = discretize(x, x_edges)
xgridclosest = xgrid(iclosest_x);
d = abs(xgridclosest-x)
Do the same for y, then
LIA = false(length(ygrid),length(xgrid));
LIA(sub2ind(size(LIA), iclosest_y, iclosest_x)) = true;