Universal Differential Equations (UDEs) combine terms from Ordinary Differential Equations (ODEs) with Neural Networks (NNs) to create a model that is a "black box" but should be able to produce perfect fits. This should be implemented in PyMob using the equinox package. It might require a relaxation of the requirements concerning what a model in PyMob is supposed to look like, a new inferer (probably making use of the optax package) and possibly a new solver (but the current JaxSolver might still work).

• Alter the model requirements so that a UDE model is accepted.
• Make the necessary changes to solve this model over time.
• Write an inferer to fit the model to some data.