if condition in for loop

Hello ,

I have a problem in finding the same rows between two matrices. I wrote the code as follows

function neigh_no = neigh_search(hex_center,neigh)

[~, n] =size(hex_center);
[~, n1]= size(neigh);
if(n~=n1)
    neigh=neigh';
end
neigh_no=[];
for i=1:length(hex_center)
    for j=1:length(neigh)
        if hex_center(i,:)== neigh(j,:)
           neigh_no=[neigh_no i];
        end
    end
end

end

Suppose if the hex_center array values are

0  0
0  1.7320
0  3.4641
0  5.1961
3  0
3  1.7320
3  3.4641
3  5.1961
6  0
6  1.7320
6  3.4641
6  5.1961
9  0
9  1.7320
9  3.4641
9  5.1961
1.5  0.8660
1.5  2.5980
1.5  4.3301
4.5  0.8660
4.5  2.5980
4.5  4.3301
7.5  0.8660
7.5  2.5980
7.5  4.33012
and the neigh values are

9  5.1961
7.5  6.0621
6  5.1961
6  3.4641
7.5  2.5980
9  3.4641

if you observe that the rows 11,12,15,16 and 24 in hex_center matrix has same rows in the second matrix. But in the output it displays only the rows 11 and 15.Can you please help me? its so important> I know its simple. But I dont know where I did mistake.

Answer 1

[Image Analyst]

That usually doesn't work for floating point numbers, just integers and inverse powers of 2. You'll have to check for a tolerance like the code examples in the FAQ show you how to do: http://matlab.wikia.com/wiki/FAQ#Why_is_0.3_-_0.2_-_0.1_.28or_similar.29_not_equal_to_zero.3F

[Naga A]

Thanks for the help.

Should I use the following condition in the if

areEssentiallyEqual = abs(hex_center(i,:)-neigh(j,:)) < tol

Am i right?

[Image Analyst]

Yes, that's right.

Answer 2

[Star Strider]

Try this one line of code instead of the entire loop structure:

[neigh, neigh_no] = ismember(neigh, hex_center, 'rows');

Also, since ‘neigh_no’ returns one value that is 0, the documentation for ismember explains ‘The output array, Locb, contains 0 wherever A is not a member of B.’ (The Locb variable in the documentation is ‘neigh_no’ here.)

EDIT — Added link to ismember documentation.

[Naga A]

Thanks for your help.

I'm trying to use that one. But it misses some of the neighbours of that hexagonal center. Instead of 5 it produces only one neighbour.

My problem here is : I have each hexagonal coordinates and center. I need to find the neighbouring hexagons and i need to use the index of that hexagon.

[Image Analyst]

Compute the distances. I think you can use pdist() if you have the stats toolbox. Then find distances less than about 1.75 times the separation to find the immediate neighbors.

[Star Strider]

You can introduce a tolerance with the ismember function by using ismembertol (introduced in R2015a).

If you have an earlier version, consider using the newest version of round (with two arguments) to round to a specific number of decimal places. If you don’t have that version of round, you can simulate it with this function to reduce argument precision for ismember:

roundn = @(x,n) round(x*10.^n).*10.^(-n);

It rounds ‘x’ to ‘n’ places to the right of the decimal (n>0) or to the left of the decimal (n<0). For n=0 it’s the usual round function.

Either of these should solve the tolerance problem. (Your data produced the desired result with my original code, so I didn’t initially suggest ismembertol or rounding.)