elementwise If statement, then apply an equation/function only to the elements that meet the criteria

Pretty self explanitory, but couldn't find an answer after much searching. i hope there's a simple solution to this:

say I have a matrix:

x = rand(3,3)

and I want to find all values in the matrix that are greater than zero and apply some elementwise equation, and for all values that are less than zero, apply a different elementwise function. like this:

if x>0

y = x.^2

elseif x<0

y = x.^3

end

is there a way to do this?

Answer 1

[Sean de Wolski]

For this specific example:

x = magic(3)-5;
idx = sign(x);
x(idx==1) = x(idx==1).^2;
x(idx==-1) = x(idx==-1).^3;

In general you can use logical indexing to turn any type of logical expression into an index.

[Alex]

you mean it should be like this, since y is the new variable:

x = magic(3)-5;

idx = sign(x);

y(idx==1) = x(idx==1).^2;

y(idx==-1) = x(idx==-1).^3;

it seems to work, but it gives me a vector for y...

[Shashank Prasanna]

make y a copy of x in the beginning and perform the same operations, it will be a matrix.

[Alex]

excellent, thank you sir!

[Alex]

hmmm again running into the same problem as below. what if none of the elements are less than zero or conversely, greater than zero? it returns an error...

[Sean de Wolski]

x = magic(3); %all values greater than zero
idx = sign(x);
x(idx==1) = x(idx==1).^2;
x(idx==-1) = x(idx==-1).^3;

%no error

I don't know what you're modifyign byut this does not return an error. The only way it would return an error is if you do not have y defined before the code runs.

you should always do this.

y = zeros(size(x));
y(etc.)

[Sean de Wolski]

So in total:

x = magic(3); %all values greater than zero
idx = sign(x);
y = zeros(size(x));
y(idx==1) = x(idx==1).^2;
y(idx==-1) = x(idx==-1).^3;

[Alex]

not sure what is going on, i think the problem is that the operations i'm using in the code are with other matrices that are the same size as the original x so it returns an error. i don't know what to do about that, because they have to be the same size.

ex:

x = magic(3); %all values greater than zero

z = magic(3)

idx = sign(x);

y = zeros(size(x));

y(idx==1) = z.*x(idx==1).^2;

y(idx==-1) = x(idx==-1).^3;

[Alex]

any ideas?

[Matt Kindig]

You could also combine both into one operation...maybe that would avoid your errors?

x = 2*rand(3,3)-1;  %some positive, some negative
idx = (x>0);  %1 where x>0, 0 otherwise
y = idx.*x.^2 + ~idx.*x.^3;

Note that x==0 will be treated just like x<0, which for this case is fine.

[Alex]

interesting solution. let me give that a try. it's a little funky for my application but if it works who cares, right?

[Alex]

that looks to have done it. many thanks to all of you. i can't believe this simple thing ended up being so difficult in the end.

Answer 2

[Azzi Abdelmalek]

x = rand(3)-0.5
y=x.^3
y(x>0)=x(x>0).^2

[Alex]

OK but what happens if there are no values of x>0 then the third line will be empty and thus generate an error

[Alex]

I did this to get around it, but if you've got a more elegant way, I'm all ears:

if x <=0

y=x.^3

else

y= x.^3

y(x>0)=x(x>0).^2

end