How does matlab determine finite-differencing intervals for optimisation algorithms such as fmincon?

Hi all,

I have been using fmincon and am aware of the options mindiffx and maxdiffx - allowing the user to set bounds on the finite difference intervals used. However, I cannot find anywhere in the documentation, or online, what logic/algorithm is used to select the finite-difference intervals? Are they fixed at the beginning, or do they change throughout an optimisation (i.e. with changing objective/constraint/variables)?

It would make sense to me that the finite difference intervals might change in size during an optimisation. Is anyone familiar with any modern methods used for this purpose? I'm aware of the method by Gill et al 1983, however, this does not seem so appropriate for constrained optimisation.

Many thanks!

[Athul Prakash]

Can you clarify the options 'mindiffx' and 'maxdiffx' ?

No mention of those in the documentation:

https://www.mathworks.com/help/optim/ug/fmincon.html

Did you mean 'DiffMinChange' / 'DiffMaxChange' ?