add encore blog post #33
Conversation
| and ordering as the input list, with each element being the similarity value for | ||
| a pair of the input ensembles. Other available measures are the clustering | ||
| ensemble similarity measure `encore.ces` and dimensionality reduction ensemble | ||
| measure `encore.dres`. |
There was a problem hiding this comment.
It would be nice to have a rule of thumb which method would fit which use-case. Links to papers or the implmentation paper chapter would also be fine.
| as | ||
| [MDAnalysis.analysis.encore](http://www.mdanalysis.org/mdanalysis/documentation_pages/analysis/encore.html). | ||
| It implements a variety of techniques for calculating similarities between | ||
| structural ensembles (trajectories), as described in this publication: |
There was a problem hiding this comment.
I think we should do a list of all that encore adds. That would include clustering. Anything else?
There was a problem hiding this comment.
There is also the PCA. Any other dimension reduction algorithms?
There was a problem hiding this comment.
The bootstraping is also good to know. But that requires better docs that should explain what it does and where the limitations are.
There was a problem hiding this comment.
A nicer link is http://devdocs.mdanalysis.org/documentation_pages/analysis/encore.html
This can be easily changed to the release docs when necessary with s/devdocs/docs/g...
There was a problem hiding this comment.
Thanks for the work and the comments - I'm reviewing the blog post. @kain88-de, for the bootstrapping, guess you mean better docs in the docs of bootstrapping itself rather than the blog post?
There was a problem hiding this comment.
guess you mean better docs in the docs of bootstrapping itself rather than the blog post
How about both? The docs need some work for sure. But I think it should also be introduced here since it is a powerful method to estimate errors when sampling is low (and more sampling can be quite expensive for MD)
| >>> print cluster_collection | ||
| 0 (size:5,centroid:1): array([ 0, 1, 2, 3, 98]) | ||
| 1 (size:5,centroid:6): array([4, 5, 6, 7, 8]) | ||
| 2 (size:7,centroid:12): array([ 9, 10, 11, 12, 13, 14, 15]) |
There was a problem hiding this comment.
The blog doesn't really explan the output. It suggests I could infer from it to which trajectory a centroid belongs to.
| from the two universes defined above, we can write: | ||
|
|
||
| ```python | ||
| >>> cluster_collection = encore.cluster([u1,u2]) |
There was a problem hiding this comment.
What clustering algorithm is chosen? Can I change it?
| In addition to standard cluster membership information, the `cluster_collection` | ||
| output keep track of the origin of each conformation, so you check how the | ||
| different trajectories are represented in each cluster. For further details, see | ||
| the documentation of the individual functions within Encore. |
There was a problem hiding this comment.
This should link to the exact docs.
|
Do you want to remove the exensive Encore part of the 0.16.0 blog post PR #26 or do you see this as first introduce features and then show them again in the release blog post? |
And yes, absolutely in favour, see #32 – blog post for the new topology system. |
| as | ||
| [MDAnalysis.analysis.encore](http://www.mdanalysis.org/mdanalysis/documentation_pages/analysis/encore.html). | ||
| It implements a variety of techniques for calculating similarities between | ||
| structural ensembles (trajectories), as described in this publication: |
There was a problem hiding this comment.
A nicer link is http://devdocs.mdanalysis.org/documentation_pages/analysis/encore.html
This can be easily changed to the release docs when necessary with s/devdocs/docs/g...
I wanted to cut that from the release blog post. I noticed that the release post is getting very long. So having this small series of blog posts about the big features is nice. In the release post I would just list them again in the beginning and link to the corresponding blog post. |
|
Yeah splitting the release post into highlights and links to things like this is a good idea imo |
| Ensemble Similarity [`encore.hes`][hes] measure: | ||
|
|
||
| ```python | ||
| >>> hes_similarities, details = encore.hes([u1, u2]) |
There was a problem hiding this comment.
what is actually stored in the details?
|
Hi,
how can we do this? I'm using the encore-post branch. Thanks! |
|
Thanks, it worked now. |
|
@mtiberti thanks. That looks a lot better already and gives nice context to the functions. The description for the similiarity measures are also good for the documentation with minimal changes. |
|
@kain88-de and @mtiberti you should sign the post; I think then it is good to go live. |
Since the next release will be pretty big I thought we could introduce some features already in blog posts. @richardjgowers started this already with mmtf (actually I got the idea from that). This post contains the encore part for the next release.
I reworked it a little bit. It would be nice if @wouterboomsma and @mtiberti could have a look over it.