Re: Custom jacobian (dfdp) in nonlin_curvefit - calling parameters?

[Pavel Hofman]

Hi Olaf,

Dne 17. 11. 19 v 7:52 Olaf Till napsal(a):
> On Sat, Nov 16, 2019 at 08:24:26PM +0100, Pavel Hofman wrote:
> > I just wonder how the 2D case should be handled.
>
> The Jacobian function is called with the 'x' reshaped to a column
> vector (like 'x(:)' would do it). The output should be calculated for
> this column vector (one row for each element of the column vector and
> one column for each partial derivative).


Now I have for f (using the default real finite differences):

f = @(p, x) ... where x is one row of indeps, and f returns two rows, 
each row for one equation (each equation uses all p elements). This code 
works OK.

If I wanted to use explicit jacobian to improve precision and 
calculation stability, please how should I handle the two equations in f 
and dfdp?

Something like this? The original row of x would be transformed to 
column twice the length, second half duplicate of the first one.
f output = column vector where first half are outputs of the first 
equation (for the first duplicate of the original x), and the second 
half are outputs of the second equation (the the second duplicate of the 
original x).

Otherwise I do not know how to convert outputs of two equations to a 
single column vector of the same length as the indeps vector.

I very much appreciate your help.

Best regards,

Pavel.