Filter matrix rows based on value of function in Matlab

This seems like a simple question but I have been unable to find an answer anywhere. If I have a Matlab matrix "A" consisting of an arbitrary number of rows, how would I filter these rows based on the value of some function "f" (the argument of which is a row vector)? In other words, how would I keep only the rows of matrix A for which f is true? I tried

    A(f(A(:)), :)

but to no success. Any help would be greatly appreciated.

So for example, if I have the matrix:

 A =

     1     1
     1     2
     1     3
     1     4
     2     1
     2     2
     2     3
     2     4
     3     1
     3     2
     3     3
     3     4
     4     1
     4     2
     4     3
     4     4

And a function f that returns true only for inputs [1 2], [1 3], [2 4] and [3 4], my desired output would be:

 ans =

     1     2
     1     3
     2     4
     3     4

[yair suari]

seems like ismember will do the work

Answer 1

[Mark Whirdy]

Shown below using an anonymous function for neatness, but the trick is just to make sure that your function (anonymous, nested, sub, or separate) returns a LOGICAL VECTOR (or a vector of linear indexes)

fn = @(x)(x>1); 
a = randn(20,1);
fn(a); % this would return a logical vector if you ran it
myfilteredVector = a(fn(a));

If I've misunderstood, could you give me an example with an input & a required output.

------

So the bones of the required operation is actually a row-wise matching? I do this by transforming each row into a single unique element by 10*A(:,1) + A(:,2), as below. Then ismember() with its mex engine can then be put to work on matching individual element. There are other matching/indexing approaches though but the primary aim is to have the function outputting a logical vector.

A = [...
     1,     1;
     1,     2;
     1,     3;
     1,     4;
     2,     1;
     2,     2;
     2,     3;
     2,     4;
     3,     1;
     3,     2;
     3,     3;
     3,     4;
     4,     1;
     4,     2;
     4,     3;
     4,     4;];
fn = @(x)(ismember(x,[12;13;24;34]));
% fn(10*A(:,1) + A(:,2)) would return a logical  vector
A(fn(10*A(:,1) + A(:,2)),:); % logical vector selects rows, & we want every column so (logic, :)

You can of course also specify the matchset a=[12;13;24;34] as an argument as below.

fn = @(x,a)(ismember(x,a));
a = [12;13;24;34];
A(fn(10*A(:,1) + A(:,2)),:);

To be honest, your function is kind of unnecessary really (unless there's a reason outside of the example you've given), as you can just use ismember with the unique encoder 10*A(:,1)+A(:,2).

a = [1,2;1,3;2,4;3,4];
A( ismember(10*A(:,1)+A(:,2), 10*a(:,1)+a(:,2)) , :)

[Mark Whirdy]

does this answer your question?

Answer 2

[Azzi Abdelmalek]

n=size(A,1);
A(arrayfun(@(x) expression(A(x,:)),(1:n)','un',0))=[]

[ARS]

Thanks for the answer. Where would the hypothetical function f come into play here?

[Azzi Abdelmalek]

expression(A(x,:))

%or

f(A(x,:))

but you have to specify your logical expression, (it's not a function)

[Cedric]

expression is what you would replace with your function name. Note that if your function returns true for rows that you want to keep, you'll want ~expression(A(x,:))

[ARS]

Sorry if I'm misunderstanding, but I receive the error:

Error using subsindex
Function 'subsindex' is not defined for values of class 'cell'.

[Azzi Abdelmalek]

Try

n=size(A,1)
A(cell2mat(arrayfun(@(x) f(A(x,:)),(1:n)','un',0)),:)=[]

Also this is not a general solution, each expression has to be treated individually.