Feature Request: Perfect Foresight Solver in GEconPy

[julia-m-schmidt]

Hi Jesse, I. would love to use a perfect foresight solver within GEconPy to simulate deterministic transition dynamics (e.g., after a policy shock or structural change).

Looking forward to testing it,
Julia

[jessegrabowski]

Hi @julia-reisach !

Sorry for not responding right away. I agree this is an important feature. It should be easy to do, because we have the symbolic graph, so one just needs to compute the relevant Jacobian matrices and block-concatenate them.

In a perfect world I'd want access to sparse jacobians, because the resulting matrix is going to be really big. I made an issue upstream in pytensor about this: pymc-devs/pytensor#1499. In gEconpy though, we have access to a lot more information about the system, so we can probably leverage that info to do something smart.

I'll be working on it in the next few weeks and I'll tag you on the eventual PR so you can check if it works on your models :)