Since I likely won't get to finish development of this so I can focus on MBAR/OST/ other things, I figured I'd push it as is. All of the communication and "resampling" is set up to run the calculations, and it conserves weight (a.k.a. - applies correct deletion/split decisions s.t. the sum of all weights = 1 as expected). I could see this being used with GK to study stability of protein/dna folds over "long time scale" simulations. It could also be used to improve crystal sampling or potentially see crystal formation.
While the important machinery works, there is still many smaller tasks that need to be improved/implemented:
Resampling priorities (who should get split and who should get merged) - there are several parameters that are hard-coded and can be modified to improve the sampling benefits seen (this is fairly simple but requires a lot of resources to run simulations to test)
Implementation of more metrics and implementation of Voronoi binning (both should be very straightforward)
Implementation of 2d & Nd metrics (2d straightforward, Nd if I remember correctly is difficult or impossible without substantial work for the user)
Improved file-reading & restarting since currently copies files over which does not scale well with trajectory length (this may be not difficult but will probably require a lot of copies flying around)
Since I likely won't get to finish development of this so I can focus on MBAR/OST/ other things, I figured I'd push it as is. All of the communication and "resampling" is set up to run the calculations, and it conserves weight (a.k.a. - applies correct deletion/split decisions s.t. the sum of all weights = 1 as expected). I could see this being used with GK to study stability of protein/dna folds over "long time scale" simulations. It could also be used to improve crystal sampling or potentially see crystal formation.
While the important machinery works, there is still many smaller tasks that need to be improved/implemented: