How can I delete 10% random Selected Index from array without changing shape in Matlab

Hello everyone, I hope you are doing well.

I have the dataset of 250x1000

I want to deleted 10% random Selected Index from array and shape will be like as input dataset. I have impleted the following code But it changes the dataset shape

Please can anybody help me in that

dataset1=dataset;
N = numel(dataset1) ; % total number of elements 
idx = randsample(N,round(10/100*N)) ;  % select 10% of indices randomly 
dataset1(idx) =[] ;   

Answer 1

[Arif Hoq]

try this:

A=load('datasetvalue.mat');
AA=A.dataset  ;
N = numel(AA) ; % total number of elements 
tenpercent=round(10/100*N);  % 10 percent of total elements
array=randi(10,250,100); % 10 percent random data
[idx]=find(array>0);
AA(idx)=[];
output=reshape(AA,250,[]);

[Med Future]

@Arif Hoq it changes the shape of the array to 250x900

[Arif Hoq]

you are deleting data, so the array size will be changed. you can replace your data if you want the same array size.

[Med Future]

@Arif Hoq not any way to solve this problem?

[Arif Hoq]

you can replace your data with 0 as @DGM suggested in his approach.

A=load('datasetvalue.mat');
AA=A.dataset  ;
N = numel(AA) ; % total number of elements 
tenpercent=round(10/100*N);  % 10 percent of total elements
array=randi(10,250,100); % 10 percent random data
[idx]=find(array>0);
AA(idx)=0;

[Med Future]

@Arif Hoq your code does not gives random idx it gives only first 100 to zero

[DGM]

Note that this only deletes the first 10% of indices, not a random sample of indices.

Also, since the array is reshaped, there is no longer any correspondence between elements across rows.

To replace (any) random element with (e.g.) NaN:

A=load('datasetvalue.mat');
AA=A.dataset;
N = numel(AA); % total number of elements 

% doing it this way makes sure the indices are unique
% otherwise you'll end up with <10% being replaced
idx = 1:N;
idx = idx(randperm(N));
AA(idx(1:round(0.1*N))) = NaN;

nnz(isnan(AA))

[Med Future]

@DGM Okay if we delete random sample then append zero at the end to comple 1000 sample per row how can i do that?

Answer 2

[DGM]

Numeric arrays cannot have "holes" in them. If you want to omit data, you either have to replace it with some value that prevents it from being influencing your process (e.g. 0 or NaN), or you need to adapt your process to address only the valid parts of your array (e.g. using a logical mask).

[Med Future]

@DGM then How this problem should be solved?

[DGM]

That depends on what you want, the meaning of your data, and how you intend to process it.

If you want the array to stay 250x1000, then you'll need to either replace the elements as mentioned or alter the process to address only the relevant elements.

If your data is essentially tabular, where each row represents a set of corresponding values, then omitting any one element in the row may make it unusable. So then, would the goal be to remove 10% of rows?

Since I don't know what your data means, or what you're doing with it, or why you need to omit values, I don't know what you need to do.