error when call function

hello,

I use matlab for excercises on the university. I declare a function 'benadering_four' and when I ask for benadering_four(0) (so 0 put in the function) then it works but when I put a vector in the function (in the code, benadering_four(xx) ) then it gives me the error:

*Error using * Inner matrix dimensions must agree.

Error in @(y)[1,cos([1:floor(n/2)]*y),sin([1:floor(n/2)]*y)]

Error in @(y)Fourier(pi*y)*coef_four*

xx = -1:0.001:1; 
n = 9;
coef_four = 1/pi*[pi*atan(5)/5; 1.08104; .552133; .30444; .157123; 0; 0; 0; 0] 
Fourier = @(y) [1 cos([1:floor(n/2)]*y) sin([1:floor(n/2)]*y)]
benadering_four = @(y) Fourier(pi*y)*coef_four
benadering_four(0)
benadering_four(xx)

I doesn't understand why it works with an number and why it doesn't work with a vector. Can someone help me?

Answer 1

[Adam]

cos([1:floor(n/2)]*y

will only be valid syntax for y being a vector in very specific circumstances - i.e. if they are the same length and one is a row vector, the other is a column vector. Depending which way round they are in that case you either get a scalar result or a square matrix result.

In all other cases you get a syntax error because you cannot do matrix multiplication (or indeed element-wise multiplication) on two arbitrary vectors of different lengths.

In this case I don't know what your expectations would be from passing a vector in. Do you expect a 3-element vector out? If so then y needs to be a column vector of length n.

[Koen]

coef_four = 1/pi*[pi*atan(5)/5; 1.08104; .552133; .30444; .157123; 0; 0; 0; 0] 
Fourier = @(y) [1 cos([1:floor(n/2)]*y) sin([1:floor(n/2)]*y)]
benadering_four = @(y) Fourier(pi*y)*coef_four

the size of Fourier is (1,9) (because floor(9/2)=4 ) and the size of coef_four is (9,1). So the vector product is well defined. benadering_four(0) works (it do give me the benadering_four function calculated in 0) but when I do benadering_four(xx) (here xx = -1:0.001:1;) i expect a vector of the same size of xx but I just get an error. (I think i don't understand your explanation)

[Adam]

Well if floor(n/2) is 4 then [1:floor(n/2)] is going to be a length 4 vector. You can't multiply a length 4 vector by whatever the length of xx is (2001 in this case).

The size of coef_four is not the problem, the error is in the Fourier part of the function.

[Koen]

I thought I would get the following: a matrix A of size (2001, 9) (2001 is the size of xx) with

A =[1 cos(xx(1)) cos(2*xx(1)) ... sin(4*xx(1));
   1 cos(xx(2)) cos(2*xx(2)) ... sin(4*xx(2));
    ...
    1 cos(xx(2001)) cos(2*xx(2001)) ... sin(4*xx(2001))]

How do I get this?

[Adam]

Try transposing xx or whatever you are passing in.

You can do matrix multiplication if you have a row vector and column vector but not two of the same.