fsolve with cell input?

As a continuation of an earlier question, I am now trying to find the minimum of a function f from R^n to R^n. I have figured out a way to make this function flexible in terms of n (an anonymous function which essentially takes in a vector input, converts the vector to a cell array, and has f evaluate that cell array), but when I minimize this anonymous function using fsolve, it is slow, due to all of the calls to fsolve, each of which call the function mat2cell. Furthermore, fsolve can apparently not take as input a cell array. Is it possible to eliminate the need for mat2cell (and the anonymous function) and maintain the complete flexibility of my code?

[Sean de Wolski]

You'd have to show us the relevant code.

[Walter Roberson]

Are you sure that f is R^n to R^n ? It is difficult to define what it means to minimize a function whose range is not R^1 .

Answer 1

[Sean de Wolski]

Using num2cell if you're only concatenating along one dimension is significantly faster. E.g:

A =  magic(1000);
t1 = 0;
t2 = 0;
for ii = 1:50
  tic
  B = mat2cell(A,ones(1,1000),1000);
  t1 = t1+toc;
  tic
  C = num2cell(A,2);   
  t2 = t2+toc;
end    
isequal(B,C)
t1/t2

I'm seeing about a 15% increase in speed with this matrix.