Let us try to keep a rough log of the changes we did.
09.01.24
- data: cleanup, refactor and streamline the api
- iso2: enable strat. sampling for diala2, fix ISA usage (no feasiblize! as it assumes first component to be constant 1)
08.01.24
- updated adaptive sampling to work in higher dimensions
- cleaned up the repo
07.12.23
- isokann was working, we had beautiful muller brown plots
- i changed a lot afterwards to glue it to the dipeptide, should test if mb is still converging
22.2.23
- fix the ominous u0 bug where all trajectories always started in the same point
- added a chifix plot
21.2.23
- use the avogadro pdb with connections (important!)
- switch to ISOSim type
- added some pairwisedists implementation fighting zygote
- comment out SimpleChains since their package precompilation was broken
- pick better T, dt defaults depending on gamma
16.2.23
- Added
scatter_ramachandran()to visualize the learned chi values over the dihedrals - Switched
defaultmodel()to use pairwise atom dists as first layer input - Switched optimiser to use L2-regularization (
OptimizerChain(WeightDecay(), Adam()))
15.2.23
- export trajectories for visualization (
extractdata(), exportdata(), savecoords()) - added Alanine dipeptide
PDB_ACEMD() - switched to use the type
MollySDEto encapsulate all integration parameters