Christian Schwantes
Christian Schwantes
I think a different test case is easiest. Right now I fixed the issues with `msmbuilder.metrics.Positions` which used LPRMSD, but I haven't replaced the LPRMSD metric yet
What do we gain from using pandas.Series? Is it just the timestamp?
Could we instead do a single pandas array with: traj, frame, time, assignment? Then we don't need a list of them
You can have a data frame with arbitrary columns. So could we just store a single data frame where each entry has time stamp trajectory and assignment
Our intention was to just make everything zero-indexed. The residue ID entry from the trajectories is taken directly from the PDB file that you specify to make it. We should...
Yea most definitely, I think what I had in mind was the geometry subpackage would essentially become the feature/representation/etc. class. On Tue, Feb 26, 2013 at 2:48 PM, Robert McGibbon...
And yea the mahalanobis would still be there. Essentially metric would be either Vectorized or RMSD. Good point on the Project. That makes sense. We would also need to decide...
We could even have a default representation (Null / atom positions) that you can set within ProjectInfo.yaml (or with ConvertDataToHDF) so that we can eliminate that from the command line...
Hmm I like that, would we then just have some file like: ``` # msmb_config.py msmbuilder.Project.default_representation = ``` I wonder, can you specify classes on the right side too? Does...
I think we can at least check for the really obvious things, like NaN's and infinities