Sparsity pattern for Jacobian and Hessian matrices

[a-jp]

Hi,

I have not used the library yet, but I'm strongly considering porting my code which uses CppAD to use Clad. With CppAD it is possible to obtain sparsity patterns when computing Jacobian and Hessian matrices - entries which are truly zero. Is this possible with Clad and if so can you point me to any documentation or preferably examples?

Thanks

[a-jp]

Or...are the jacobians and hessians that are returned by clad, done so only containing non-zero entries? They are already sparse? If so, how does one infer the location of the zeros when rebuilding a dense matrix?

Thanks

[vgvassilev]

Hi @a-jp, thanks for your question.

Currently the way we model jacobians and hessians is without taking into account sparsity. We have an issue which we will tackle soon in that respect #509. Perhaps you could help us understand your usecase by showing a small example?

[a-jp]

Hi @vgvassilev I was just following up my most recent post to see if you had any comments? Thanks

[vgvassilev]

@a-jp, apologies for the delay. I think I understand your use-case by the interfaces you described a little better now. I think we can tackle your requirement in clad v1.6 which is scheduled for end of May. Do you think you can capture the gist of this discussion in an issue?

[a-jp]

Hi, yep I'll take a first stab at that as soon as possible.

[a-jp]

Please see issue #803. Thanks

[vgvassilev]

Thank you!