How to concatenate the data in each iteration in the same variable?

I am saving the output of the 1st iteration in the images_TRO_pressing , images_TRO_slip, images_TRO_noise, the size of each variabe is 120*120*300

In the other iterations how can I keep adding the data in the same varibales so I end up with size for each of them around 120*120*3600? as well same to the varibale labels_TRO with size 1*900 for one iteration

Note that the third dimension which is the number of matrices varies in each iteration not all iterations give 300.

for i = 1:12      
pressing_labels = find(labels_all_Train1(i).labels == 0);
        slip_labels = find(labels_all_Train1(i).labels == 1);
        noise_labels = find(labels_all_Train1(i).labels == 3);
        minimum = min([length(pressing_labels), length(slip_labels), length(noise_labels)]);
        random_pressing = randsample(pressing_labels, minimum);
        random_slip = randsample(slip_labels, minimum);
        random_noise = randsample(noise_labels, minimum);
        images_TRO_pressing = HM_all_Train1(i).TRO_heatmap(:, :, random_pressing);
        images_TRO_slip =  HM_all_Train1(i).TRO_heatmap(:, :, random_slip);
        images_TRO_noise = HM_all_Train1(i).TRO_heatmap(:, :, random_noise);
        labels_TRO = [labels_all_Train1(i).labels(random_pressing), labels_all_Train1(i).labels(random_slip), labels_all_Train1(i).labels(random_noise)];

    end

[M]

Hi @Dyuman Joshi, Do you have any idea please? thanks

[Dyuman Joshi]

I'll get back to you in some time.

[M]

Hi @Dyuman Joshi , Please dont forget because I am stopped here, thanks

[Dyuman Joshi]

As for storing the data, You can preallocate a cell array to store arrays and then concatenate them after the for loop.

Also, you can use randperm instead of randsample. Stating this because the former does not require any toolbox, but the latter one does.

[M]

@Dyuman Joshi

logical indexing is going to be faster.

I didnt get how this can return the index of a certin value ? doesnt it return the values itself?

you can preallocate a cell array to store arrays and then concatenate them after the for loop.

Could you please provide me with example because I tried it but it keeps only return the last iteration I dont know what is the error

you can use randperm

Can you please show me how can I use it to take a random sample with a certain number of samples from a vector, I didnt find this in the documentation

I appreciate your help, thanks

[Dyuman Joshi]

"Could you please provide me with example because I tried it but it keeps only return the last iteration I dont know what is the error"

Try this -

n=12;
labels_TRO = cell(n,1);

for i = 1:n
        %Allocate data which needs to be accessed frequently to a variable 
        arr = labels_all_Train1(i).labels;
        
        pressing_labels = nnz(arr == 0);
        slip_labels = nnz(arr == 1);
        noise_labels = nnz(arr == 3);
        
        minimum = min([pressing_labels, slip_labels, noise_labels]);
        random_pressing = randperm(pressing_labels, minimum);
        random_slip = randperm(slip_labels, minimum);
        random_noise = randperm(noise_labels, minimum);
        images_TRO_pressing = HM_all_Train1(i).TRO_heatmap(:, :, random_pressing);
        images_TRO_slip =  HM_all_Train1(i).TRO_heatmap(:, :, random_slip);
        images_TRO_noise = HM_all_Train1(i).TRO_heatmap(:, :, random_noise);
        labels_TRO{i} = [arr(random_pressing), arr(random_slip), arr(random_noise)];
end

See Preallocation for more info.

"Can you please show me how can I use it to take a random sample with a certain number of samples from a vector, I didnt find this in the documentation"

%4 random numbers from 1 to 10
randsample(10,4)

ans = 4×1

9 3 6 5

randperm(10,4)

ans = 1×4

6 7 9 1

[M]

@Dyuman Joshi these 3 lines returen to me wrong indices

 pressing_labels = nnz(arr == 0);
        slip_labels = nnz(arr == 1);
        noise_labels = nnz(arr == 3);
        

When I got back to "find"

I got the following error

Error using randperm

Inputs must be nonnegative scalar integers.

[M]

Also, as you only need the number of elements that satisfy the condition

nope, I need the indices to do matching with the images!!

See nnz doesnt return the indices!!

A = [1 0 2 3 1 2 4 3 1 4] ;
nnz(A == 0)

ans = 1

[M]

@Dyuman Joshi I tried to use the logical indexing as the following but I got this error:

  pressing_labels = arr == 0;
        slip_labels = arr == 1;
        noise_labels =arr == 3;
        

Error using randperm

Inputs must be nonnegative scalar integers.

even with randsample I got empty array here with the logical indexing

        labels_TRO{i} = [arr(random_pressing), arr(random_slip), arr(random_noise)];

[Walter Roberson]

@M

You would not do something like randperm(pressing_labels, 3) .

Instead you would do something like

P = randperm(numel(pressing_labels), 3);
rand_PL = pressing_labels(P);

[Dyuman Joshi]

"nope, I need the indices to do matching with the images!!"

Yeah, my bad. I overlooked the fact that the indices are used as input to randsample as well.

Just to confirm, you need 4 outputs (images_TRO_pressing, images_TRO_slip, images_TRO_noise and labels_TRO) with values from all iterations?

[M]

@Dyuman Joshi yes exactly

[M]

@Walter Roberson your solution didnt work, I got the same error

[Walter Roberson]

  pressing_labels = find(arr == 0);
  slip_labels = find(arr == 1);
  noise_labels = find(arr == 3);
  
  N = 4;   %number to be sampled
  if numel(pressing_labels) < N; error('not enough pressing locations to sample'); end
  if numel(slip_labels) < N; error('not enough slip locations to sample'); end
  if numel(noise_labels) < N; error('not enough noise locations to sampe'); end
  
  random_pressing = pressing_labels(randperm(numel(pressing_labels), N));
  random_slip = slip_labels(randperm(numel(slip_labels), N));
  random_noise = noise_labels(randperm(numel(noise_labels), N));

However you do not want to take arr(random_pressing) and so on -- you know that arr() at pressing locations is going to be 0 because that is how you chose those locations, according to what was in arr() there.

[M]

@Walter Roberson Thanks a lot, this is worked and thanks for @Dyuman Joshi

but I am wondering why the logical indexing instead of "find" didnt work, any suggestion please?

[Walter Roberson]

You have not been clear as to what your code is, or what your desired outcomes are.

If you are trying to find N locations each of which correspond to a particular arr value, then you have a few possibilities:

• do NOT just take arr(LOGICAL MASK) since you already know what the arr values are at those locations because that is how you chose the locations
• taking a random sub-sampling of a logical mask would get you are random number of elements selected, and would not tell you where they were selected from
• using find() to locate the elements and taking a random selection from those indices works fine if what you want to know is the locations
• You could create a vector 1:numel(arr) and index that vector at a logical matrix such as arr==0 ... but you would effectively just have duplicated the work of find() less efficiently