[scilab-Users] optimizing non-differentiable functions

[Eric Dubois]

If your problem is only that you cannot calculate the deriative (and not
that it does not exist), then you can use a numerical derivative (such as
the Scilab function derivative). To that end, if your function is func, then
create a function newf:

function [f,g,ind] =newf(param,ind)
f=func(param)
g=derivative(func,param)
endfunction


2009/8/16 Sebastian Urban <urban at stoch.uni-karlsruhe.de>

> Hello everybody,
>
> is there a way to optimize a function, whose gradient I don't know /
> cannot compute? The command 'optim' seems to require the cost function
> to provide the gradient...
>
> Thanks for help, suggestions and comments!
> Sebastian
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20090818/ef447665/attachment.htm>