Making a group of maximum nearest elements.

If I have some points and a data. I need to find the all the nearest elements of points. How can I do that ? Please help me.

[Guillaume]

What's your definition of nearest elements?

[BHAGYALAKSHMI M]

The nearest elements that are close to each point. I atttached a data here.

Suppose I have some elements A= [1,4,2,6,9,0,4.2,5.6,....] and points

B=[0.1,4.1,3.......]. Calculating and comparing the distance of these and based on that grouping.

Taking point 1 from A checking is this element is close to 0.1 or 4.1. It is close to 1 more than 4 so

G1=[1,0.1].

Taking point 4. It is close to 4.1, so

G2=[4,4.1].

2 is close to 3, so

G3=[2,3].

6 is close to 4.1 compared to others, so

G2=[4,4.1].

The process repeats until no elements in A.

How can I do that? Please help me.

Answer 1

[Guillaume]

Be aware that the following creates a temporary matrix of size numel(A) x numel(B), so if both vectors are very large you may run out of memory:

%for row vectors:
%example data
A = [1,4,2,6,9,0,4.2,5.6];
B = [0.1,4.1,3];

assert(isrow(A) && isrow(B), 'Inputs must be row vectors');
[~, groupA] = min(abs(A - B.'), [], 1)

%for column vectors:
%demo data
A = A.'; B = B.';

assert(iscolumn(A) && iscolumn(B), 'Inputs must be column vectors')
[~, groupA] = min(abs(A.' - B), [], 1)

I'm not sure how this should be applied to your example datasheet. Looking at what it contains, you should import that excel file in a table, in which case you should store the group as another variable of the table so you can then use aggregation functions such as groupsummary.

[BHAGYALAKSHMI M]

Thank you sir.

I am not getting the output,there was some errors.

The following is my points.

B=
Columns 1 through 10

   44.8481    1.7293    1.4514    0.6036  199.6823   29.9475   35.4337    6.5257    3.2450    0.9861
   39.3333    1.8333    1.6500    0.6500  492.5000   26.8333   32.1667    6.5333    2.6167    0.6617
   57.4286    1.7143    2.0714    0.9286  198.0000   53.1429   93.4286    5.8000    2.9000    1.0114
   47.0000    2.0000    9.0250    4.5750  207.0000   62.2500  181.2500    6.7500    2.7000    1.1750
   57.0000    2.0000    1.4000    0.7000  470.0000   62.0000   88.0000    5.6000    2.5000    0.8000
   23.3333    2.0000    3.4333    1.4667  287.3333   67.3333  139.3333    5.6000    2.4667    0.7667
   11.0000    1.3333    0.8333    0.1667  392.0000   25.0000   66.3333    7.0667    3.5667    1.0667
   65.0000    2.0000    0.7000    0.1000  392.0000   20.0000   30.0000    5.3000    2.8000    1.1000
   54.0000    1.0000   22.6000   11.4000  558.0000   30.0000   37.0000    7.8000    3.4000    0.8000
   34.0000    2.0000    3.7000    2.1000  490.0000  115.0000   91.0000    6.5000    2.8000    0.7000
   70.0000    2.0000    2.7000    1.2000  365.0000   62.0000   55.0000    6.0000    2.4000    0.6000
   54.3333    2.0000    5.1000    2.2000  223.3333   60.3333  149.0000    6.9000    2.8333    0.6333

  Column 11

    1.3867
    1.5000
    1.2857
    1.0000
    1.0000
    1.0000
    1.0000
    1.0000
    1.0000
    1.0000
    1.0000
    1.0000

How can I find nearest points of this from my dataset? Please help me.

[BHAGYALAKSHMI M]

I am getting an error

"Error using wn (line 25)

Inputs must be row vectors"

How can I get the result?

[Guillaume]

Well, yes I explicitly wrote the check that the inputs were vectors. Why is the input a matrix? What is the desired result when the input is a matrix? Are both A and B matrices? What is to be checked againt what? etc.

You need to give a lot more explanation as to what exactly you want. An explanation of what the inputs represent would also help.

[BHAGYALAKSHMI M]

If I take the unique values, then I get some B like--

B=0.1000
    0.1667
    0.6000
    0.6036
    0.6333
    0.6500
    0.6617
    0.7000
    0.7667
    0.8000
    0.8333
    0.9286
    0.9861
    1.0000
    1.0114
    1.0667
    1.1000
    1.1750
    1.2000
    1.2857
    1.3333
    1.3867
    1.4000
    1.4514
    1.4667
    1.5000
    1.6500
    1.7143
    1.7293
    1.8333
    2.0000
    2.0714
    2.1000
    2.2000
    2.4000
    2.4667
    2.5000
    2.6167
    2.7000
    2.8000
    2.8333
    2.9000
    3.2450
    3.4000
    3.4333
    3.5667
    3.7000
    4.5750
    5.1000
    5.3000
    5.6000
    5.6000
    5.8000
    6.0000
    6.5000
    6.5257
    6.5333
    6.7500
    6.9000
    7.0667
    7.8000
    9.0250
   11.0000
   11.4000
   20.0000
   22.6000
   23.3333
   25.0000
   26.8333
   29.9475
   30.0000
   32.1667
   34.0000
   35.4337
   37.0000
   39.3333
   44.8481
   47.0000
   53.1429
   54.0000
   54.3333
   55.0000
   57.0000
   57.4286
   60.3333
   62.0000
   62.2500
   65.0000
   66.3333
   67.3333
   70.0000
   88.0000
   91.0000
   93.4286
  115.0000
  139.3333
  149.0000
  181.2500
  198.0000
  199.6823
  207.0000
  223.3333
  287.3333
  365.0000
  392.0000
  470.0000
  490.0000
  492.5000
  558.0000

Choosing this as B how can I find all neighbors of these points.

A is a data, B is some points suppose B contains 50 points then we will get 50 groups. Suppose the data contains some points like: {1,0.3,4,7,9,23,55, etc} (a matrix). B contains:{ 0.4,10,44,60,etc}.

Check the whole data, the first point is '1' in the data. This point is close to 0.4 compared to the other points in B. Next is 0.3, this also close to 0.4 compared to others. So we got a group like 0.4,0.3,1. This process continues when all the elements in the data checked. At last, we get some groups/clusters. Like

C1=[1,0.4,0.3,4]

Then 7 (from data) that is close to 10 then

C2=[7,10,9,23]

55 close to 60 compared to others

C3=[55,60]

The 44 is remaing, this is close to C3 elements so

C3=[44,55,60].

How can get some groups that contains(the groups) all the elements inthe dataset ?

Please help me.

[BHAGYALAKSHMI M]

Is there any method to find this?

[Guillaume]

If B is a vector and A is a matrix and you want to find the index of the element of B each element of the A matrix is closest to (= your group number), then:

assert(isvector(B), 'B must be a vector');
[~, groupnumber] = min(abs(reshape(A, 1, []) - B(:)), [], 1);
groupnumber = reshape(groupnumber, size(A));

This gives you a matrix the same size as A indicating which group each elements falls in.

Be aware that this temporarily creates a numel(A) x numel(B) matrix. For very large matrices, you may run out of memory.

[BHAGYALAKSHMI M]

Thank you sir. From this how can I get Group1 elements?

[Guillaume]

A(groupnumber == 1)

is all the elements of group 1. I'm not sure what you're planning to do afterwards. Be aware that creating a matrix for each group is not a good approach. There are plenty of matlab functions that will let you apply the same algorithm to all the groups at once, as long as everything is kept together in one matrix.

For example, if you want to calculate the mean of each group in A:

groupmean = groupsummary(A(:), groupnumber(:), 'mean'); %calculate the mean of each group of A

[BHAGYALAKSHMI M]

I want to take each group elements. Like:

group(1)=[elements]

group(2)=[elements] etc.

How can I get this?

[Guillaume]

As I've said, this is typically a bad idea. I was asking about what you want to do after that. It is very likely that whatever it is, it will be much easier if you don't do this.

Now if you really insist:

group = accumarray(a(:), groupnumber(:), [], @(v) {v});

or

group = splitapply(@(v) {v}, a(:), groupnumber(:));