Open axsk opened 1 week ago
We have a first draft in https://github.com/axsk/ISOKANN.jl/commit/593178fdca2503b585bc4a1a4575318d16a046e7
I am not sure if the units/magnitudes are correct yet.
Whereas the integrator should work (still untested), we also need a way to expose it to propagate etc.
I currently think of an alternative OpenMMSimulationControlled
that has an additional force::Function
field.
@joramkuntze: keeping you in the loop :)
It should be working now for the overdamped case, if dt/gamma <= 1e-6
Unfortunately OD isn't directly comparable to UD, so maybe we should implement this now for UD Langevin. In principle it works the same, just the update step is a little more complicated, cf https://arxiv.org/abs/2303.14696 (eqs 23 and table I) or https://pubs.acs.org/doi/full/10.1021/acs.jpcb.4c01702 (eq. 16,17 or table I)
For implementing the optimal control from our article.