[ot-users] AbdoRackwitz steps

Anita Laera anita.laera87 at gmail.com
Wed May 24 16:06:12 CEST 2017


I am performing FORM analysis with AbdoRackwitz optimization algorithm.

I am using:

gradient = ot.CenteredFiniteDifferenceGradient(steps, f.getEvaluation())

(where f is a NumericalMathFunction())

steps = ot.ConstantStep(epsilon)

I tried two different values of epsilon: 0.00001 and 1.0.

I couldn’t see any difference in the input parameters that are considered
at each evaluation. Why?

How is the value of epsilon going to affect the input parameters?

Also, after the first evaluation, the message:

‘WRN - You are using a default implementation for the gradient. Be careful,
your computation can be severely wrong!’

is printed. Is it because I am not defining the gradient through the
_gradient(self, X) method?

Thanks in advance for your help.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openturns.org/pipermail/users/attachments/20170524/32fe5515/attachment.html>

More information about the users mailing list