choose 2 random unique element inside cell

how to choose 2 random unique element inside a cell?

index={[1,2,5,9,10,13,17,18,21],[4,5,7,12,13,15,18,20,21],[3,4,6,11,12,14,18,19,20],[8,16,20,22]};

example of result

p={[1,9],[7,21],[3,4],[16,8]}

Answer 1

[per isakson]

See datasample, Randomly sample from data, with or without replacement

and try

>> cellfun( @(row) datasample( row, 2, 'replace',false ), index, 'uni',false )
ans =
  1×4 cell array
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}

"[...] random unique element inside a cell" does that mean without replacement?

If it means literal "unique elements" then run this is a first step to remove duplicates.

>> index = cellfun( @unique, index, 'uni',false )
ix =
  1×4 cell array
    {1×9 double}    {1×9 double}    {1×9 double}    {1×4 double}

Or should the propability to pick a specific value be proportional to the number of duplicates of that value?

[NA]

As I want to run simulation 5 times, how can I fix datasample?

[per isakson]

Problem: You have a working script that runs the simulation once and you want to run it automatically many times

The first approach you shall consider is a for-loop.

1. Convert the script to a function
2. Create a second function that calls the first many times
3. Run the second function

Example

>> out = cssm( index, 5 )
out =
  5×1 cell array
    {1×4 cell}
    {1×4 cell}
    {1×4 cell}
    {1×4 cell}
    {1×4 cell}
>> out{3}
ans =
  1×4 cell array
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}

where

function  out = cssm( index, n )  
    out = cell( n, 1 );
    for jj = 1 : n
        out{jj} = run_once( index );
    end
end
function  out = run_once( index )
    out = cellfun( @(row) datasample( row, 2, 'replace',false ) ...
                  , index, 'uni',false );
end

BTW:

%%
index = repmat( index, 5,1 );
out = cellfun( @(row) datasample( row, 2, 'replace',false ), index, 'uni',false )

outputs

out =
  5×4 cell array
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}
    {1×2 double}    {1×2 double}    {1×2 double}    {1×2 double}

[per isakson]

What happened to the comments on the interpretation of "unique"?

[NA]

I have this code and want to run it 5 times (amount is changing in each iteration)

for iteration=1:5
    index={[1,2,5,9,10,13,17,18,21],[4,5,7,12,13,15,18,20,21],[3,4,6,11,12,14,18,19,20],[8,16,20,22]};
    amount={[10,11,12,10,11,10,12,13,20],[11,15,12,13,11,10,15,13,12],[13,16,16,17,11,10,15,13,12],[4,5,6,10]}; % amount is changing in each iteration
    changed_num=2;
    if iteration==1
        changed_mes_index=cellfun( @(m) datasample( m, changed_num, 'replace',false ), index, 'uni',false );
    end
    b=cellfun( @(t,m,u) t/2.*(ismember(m,u)), amount,index,changed_mes_index, 'uni',false );
    result=cellfun(@(m,y) m+y, amount,b, 'UniformOutput', false);
end

[per isakson]

"you mean I do something like this" No!

What happend to "I want to run simulation 5 times"?

for iteration=1:5
    ...
end

isn't meaningful. It does the same thing five times and overwrites result. How is amount supposed to change?

It's hard to help, since you don't try to understand my answer and move the goalposts in the comments.

And I repeat, what happened to the comments on the interpretation of "unique"?

[NA]

I want to do something like this

    if iteration==1
        changed_mes_index=cellfun( @(m) datasample( m, changed_num, 'replace',false ), index, 'uni',false );
    end
    

this code

    index = repmat( index, 5,1 );
out = cellfun( @(row) datasample( row, 2, 'replace',false ), index, 'uni',false )

does not give what I want

[NA]

amount is calculated by adding normrnd. so it changes in each iteration.

[per isakson]

"amount is calculated by adding normrnd" That means that you need to be careful regarding Accuracy of Floating-Point Data

So far, I have assumed whole numbers

Answer 2

[Jos (10584)]

If I understand you correctly, the final output should have 8 unique numbers, in four groups of two, where the numbers in each group is drawn from a cell in the cell array index. This is not a trivial problem to answer! There might situations where this is simply impossible given a cell array index (for instance index= {[1 2 3], [1 2 4], [2 3 4],... }

You might want to try a brute-force method

index={[1,2,5,9,10,13,17,18,21],[4,5,7,12,13,15,18,20,21],[3,4,6,11,12,14,18,19,20],[8,16,20,22]};
fn = @(v) sort(v(randperm(numel(v),2))) ; % function to draw 2 randomelements from v
k = 1 ;
while k < 1e5 % try a lot of times
    C = cellfun(fn, index,'un',0) ;
    if numel(unique([C{:}])) == 8,
        break
    else
        C = [] ; k = k+1 ;
    end
end
celldisp(C)

[Jos (10584)]

This comment has nothing to do with my answer ...