Exp/3 pca seq

[pablosmig]

Experimenting with issue #860. Non functional changes and breaks many tests.

[eurunuela]

Thank you @pablosmig! It's been fun to hack with you at Brainhack Donostia.

I'll finish the PR and share with the core developers team.

[eurunuela]

@tsalo @jbteves @handwerkerd where do you think would be a good place to save the PCA variance explained plots that have the Kundu options incorporated?

[handwerkerd]

I guess the core issue is that the kundu method isn't finding a local minimum so it's not actually a plot and there isn't even a threshold. The number of components retained would be an easy addition to desc-PCA_cross_component_metrics.json The other bit of info from kundu is which components are retained. That could either be a list in the same file or a new column with accepted & rejected in desc-PCA_metrics.tsv

Any other information that's looking for a home?

[dowdlelt]

Hopefully I'm not completely missing the point here, but this could be added to the existing plots - not as a threshold, but as maybe individual filled circles on the lines (or along the x-axis) indicating which components were selected. Such that the user can see that the three main methods would have picked threshold x, y, z and 'kundu' would have picked most of the components below that threshold and then a few beyond it.