RL algorithm implementation #19
Conversation
…le based on how many controls there are. using mesolve with QobjEvo object _H
_infidelity() more general
… train early, removed update_solver(), use of var_time = True supported, other minor fixes
There was a problem hiding this comment.
Thank you very much, @LegionAtol, for incorporating the feedback.
There are a few things left to do but I am sure you will be able to address them easily.
Please ping me, once you are done.
src/qutip_qoc/pulse_optim.py
Outdated
| @@ -84,6 +84,11 @@ def optimize_pulses( | |||
| Global steps default to 0 (no global optimization). | |||
| Can be overridden by specifying in minimizer_kwargs. | |||
|
|
|||
| - shorter_pulses : bool, optional | |||
There was a problem hiding this comment.
This is basically what the __time__ control parameter is used for.
Of course, RL doesn't really take a 'guess' (as with the other parameters), but we could use the same structure to only look for shorter pulses if the __time__ bounds are specified.
… added underscore for internal variables, args is now passed as a parameter to _infid(), changes in callback function
…m continues to search for solutions with shorter pulses
Pull Request Test Coverage Report for Build 11141454500Warning: This coverage report may be inaccurate.This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
Details
💛 - Coveralls |
|
@LegionAtol please take one last look at the minor changes I have made.
Once you sign of on these changes, the PR is ready to merge. |
|
Ok, that's fine |
@LegionAtols GSOC RL algorithm
No description provided.