search

markovmodel / pyemma_tutorials

How to analyze molecular dynamics data with PyEMMA
Creative Commons Attribution 4.0 International
71 stars 34 forks source link

Hyperparameter choice: VAMP-2 lagtime #144

Closed cwehmeyer closed 6 years ago

cwehmeyer commented 6 years ago

Section 3.2 requires an explanation why we use lagtime 0.5 ns for the VAMP-2 analysis (point to NB 00).

marscher commented 6 years ago

I guess the same applies for the used TICA lag time?

cwehmeyer commented 6 years ago

Yes, you're right.

franknoe commented 6 years ago

If we use VAMP/TICA to select features or find a relevant subspace to work in, the main idea is that this choice should be invariant wrt the lagtime within a significant lagtime range.

cwehmeyer commented 6 years ago

If we use VAMP/TICA to select features or find a relevant subspace to work in, the main idea is that this choice should be invariant wrt the lagtime within a significant lagtime range.

That's what we look at in the notebook. The manuscript, however, lacks this information.

We need to decide whether to refer more clearly to the showcase notebook or include more figures and explanation in the manuscript.

Personally, I do not like the latter as it would clutter the workflow description and distract from the bigger picture.

franknoe commented 6 years ago

Just add a sentence or two explaining this. If we add a Theory section, feature selection could be part of that and there we some space to explain.

Am 24/08/18 um 09:18 schrieb Christoph Wehmeyer:

If we use VAMP/TICA to select features or find a relevant subspace
to work in, the main idea is that this choice should be invariant
wrt the lagtime within a significant lagtime range.

That's what we look at in the notebook. The manuscript, however, lacks this information.

We need to decide whether to refer more clearly to the showcase notebook or include more figures and explanation in the manuscript.

Personally, I do not like the latter as it would clutter the workflow description and distract from the bigger picture.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/markovmodel/pyemma_tutorials/issues/144#issuecomment-415773033, or mute the thread https://github.com/notifications/unsubscribe-auth/AGMeQt4JnpxExUQnLP6YdEJHAyaM0JIZks5uUAtMgaJpZM4WK_Cd.

--

Prof. Dr. Frank Noe Head of Computational Molecular Biology group Freie Universitaet Berlin

Phone: (+49) (0)30 838 75354 Web: research.franknoe.de Mail: Arnimallee 6, 14195 Berlin, Germany