Is there a modify_params_after_update?

[guruvamsi-policharla]

Currently, there's a modify_params_after_iter option which allows injection of arbitrary code after each iteration of Krotov.
Is there by any chance an option to inject arbitrary code after each update (time step)?

[goerz]

There is not, but you can "inject code" for most use cases due to the fact that the propagate routine that you pass to optimize_pulses is under your control. Specifically, you have direct read/write access to both the propagated state and the Hamiltonian/Liouvillian.

The two specific use cases that I've come across are:

• Write out observables during the forward/backward propagation.
• Adjust a "moving grid" Hamiltonian, as e.g. in Fürst et al. New J. Phys. 16, 075007 (2014)

Both of these are very straightforward to do inside a custom propagate routine.

You could conceivably extend the scope of what propagate has access to with some Python trickery (closures), e.g. to give it access to the objectives:

def make_propagator(objectives):

    def propagator(H, state, dt, c_ops=None, backwards=False, initialize=False):
        pass
        # do the propagation, and something involving objectives

    return propagator

# optimize_pulses(objectives,..., propagator=make_propagator(objectives), ...)

For anything beyond that, you can always create your own copy of the optimize_pulses routine and modify that to your needs. I'd have to see a very convincing use case that isn't covered by one of the two approaches I outlined above to consider adding further hooks to the implementation in the krotov package.