Closed qasimpars closed 2 years ago
PCA is only unique up to the sign so this is what you are observing here. That put aside it is usually not a good idea to use PCA on time-series data. For projection purposes I recommend looking at TICA/VAMP/EDMD family of methods.
Hi,
I was doing PCA using PyEMMA-2.5.1 and GROMACS-2020.1.1 and compare their results. I just noticed that they give a bit different projection results. Is this a bug in PyEMMA?
Code used for PyEMMA:
Output:
GROMACS commands:
Output: