From: fuaada mohd siam on
Hello,

Does anyone here familiar to simulated annealing? I wanted to use this algorithm in MATLAB to estimate 6 parameters in my linear system ODE . I was using simulannealbnd but why it give different value each time i run the code. As far as i know, simulated annealing is the efficient way to get the global minimum value and it contains random number in the algoritm, as a result, the value may be different. Since i am not really familiar to this method, can someone here explain a bit further, and which result is the best since it gives different value everytime we put different initial value........

Thanks in advance!
Fuaada
From: Alan Weiss on
fuaada mohd siam wrote:
> Hello,
>
> Does anyone here familiar to simulated annealing? I wanted to use this
> algorithm in MATLAB to estimate 6 parameters in my linear system ODE . I
> was using simulannealbnd but why it give different value each time i run
> the code. As far as i know, simulated annealing is the efficient way to
> get the global minimum value and it contains random number in the
> algoritm, as a result, the value may be different. Since i am not really
> familiar to this method, can someone here explain a bit further, and
> which result is the best since it gives different value everytime we put
> different initial value........
>
> Thanks in advance!
> Fuaada

In fact, simulannealbnd is usually the least efficient way to search for
a global solution, as explained here:
http://www.mathworks.com/access/helpdesk/help/toolbox/gads/bsa_e70.html#bsbjg7m

I suggest using patternsearch, and try starting from a variety of random
initial points, along the lines described here:
http://www.mathworks.com/access/helpdesk/help/toolbox/gads/bsc5_cj.html#bsfjxr9

Alan Weiss
MATLAB mathematical toolbox documentation
From: fuaada mohd siam on
Hi Alan,

Thanks for your suggestions!...

Fuaada