Anonymous function with summation of parameters

Hello,

I would like to create an anonymous function to minimize. The problem is that I have many variables that are in a matrix where each column represents a variable and lines the numbers of observations. I do not want to write the function as:

 fun = @(b) (Y - b(1)*XX(:,1) - b(2)*XX(:,2) - b(3)*XX(:,3) - ... - b(N)*XX(:,N) );

Nobody knows how to include a type of summation because my attempts have not worked. I try this:

 N = 100;
 fun = @(b) (Y - symsum(b(t)*XX(:,t),t,1,N));

Thank you very much!

Answer 1

[Star Strider]

Unless I’m missing something, ‘fun’ involves straightforward matrix multiplication:

Y  = randi(9, 10, 1);                       % Create Data
XX = randi([-9 9], 10, 5);                  % Create Data

fun = @(b) sum(Y - XX*b);                   % Function

B0 = rand(size(XX,2),1);                    % Initial Parameter Estimates
B = fminsearch(fun, B0);                    % Estimate Parameters

EDIT — Also consider:

B = XX\Y;

[Amor-Aniss Benmoussa]

Thanks a lot!

I found the answer with:

fun = @(b) Y - XX*b;

I thought that I had to define the different parameter b(1) to b(N) but this is not necessary.

Amor

[Star Strider]

My pleasure!

That will not give you the correct result.

This will:

fun = @(b) norm(Y - XX*b);

(This also corrects an error in my original code, where I used sum instead of norm.)

The most efficient code remains:

b = XX\Y;

This will also give you the correct result, (the same as using norm in ‘fun’), and much more efficiently.