PCA: Eigenvalues and eigenvectors

[Agustin Lobo]

Is it currently possible to get eigenvectors and eigenvalues as output after fitPCA? A plot

of % of total variance accounted for each PC is important to decide how many PC to retain.

In case this is not possible, please add this feature.

[Andreas Janz]

All relevant information about any learner should be accessable via the Data Sources Panel. View the ScikitLearn docs for a comprehensive description of each PCA parameter (https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html).

Not sure how convinient it is to get the data out and create a plot externally. If that is an issue, please open a separate issue for that.

‌

[Agustin Lobo]

With Version 3.7.20201029T143643.master I cannot see this information anymore.

‌

[Andreas Janz]

@Benjamin Jakimow I guess your last changes regarding the data source tree broke this!?

[Benjamin Jakimow]

@Andreas Janz can you append an transformer model here?

‌

[Andreas Janz]

Use the create testdata PA:

‌

[Benjamin Jakimow]

Is already fixed on master, just needs to be uploaded as new version

‌

[Andreas Janz]

@Benjamin Jakimow how would a user copy the data displayed in the tooltip?

The “Copy” in the context menu is not very helpful, because it just gives me the “ndarray” string

I would suggest to copy the actual array values into the clipboard. And please make sure to really copy the values and not only the string representation with potential “…” inside!!!

‌

‌

[Benjamin Jakimow]

“copy” means writing these values into the clipboard. Do you think it is a good idea to copy potentially very large arrays into the clipboard?

[Andreas Janz]

The goal is, that the user can get his hands on the actual data.

“Copy to Clipboard” would be one option. “Save to File” could be another option. Or “Show in TextWidget”.

All three would make sense. Maybe start with “Save to File”.

[Benjamin Jakimow]

added HubFlowPyObjectTreeNode with own context menu addresses #517 addresses #430

Signed-off-by: Benjamin Jakimow [e-mail] [e-mail]

→ <<cset ac4d2a082498>>

[Andreas Janz]

resolves #517

→ <<cset aa77187f25f7>>

[Andreas Janz]

Numpy arrays are now copied to clipboard with full precision (no … filler!)