ENH: Viz for spatial filters

[Genuster]

What does this implement/fix?

Adds SpatialFilter container for visualisation of filters, patterns and eigenvalues of GED-based algorithms (CSP, SPoC, SSD, XdawnTransformer) and other linear models (LinearModel).

• implement plot_filters, plot_patterns, plot_scree
• make possible to plot inverse_transform'ed filters/patterns in case of pipeline (as get_coef does)
  • make get_coef work with arbitrary step in the pipeline (not only the last LinearModel)
• cover tests
• replace CSP's plots with internal functions
• make existing examples use the container's plotting methods

[wmvanvliet]

I'm all for adding nice visualizations for filters/patterns. Suggestion: would a MixedIn pattern make sense here instead of the class you have now?

[Genuster]

Cheers @wmvanvliet!

My original thought was to make it usable with external spatial filters (PyRiemann transformers, for example).
I also realize now if the relevant transformer is in the middle of the pipeline and I want visualize its filters/patterns after inverse-transform, it would make sense to use a plain function that would instantiate the Viz class instead of adding it as a method to GED/LinearModel.
If I go with Mixin this case would be problematic, as far as I can tell.

Does that make sense? If you have an idea how we can work around it with Mixin, I'd be glad to hear

[larsoner]

Looks like a good start! Needs some tutorial to show how it's useful though. Bonus points if you can modify some existing example to make it shorter and/or clearer

[Genuster]

@larsoner, I think it's ready for the next round!
I didn't add a tutorial specifically for the class, but used most of the functionality across the examples, hope it's enough

[larsoner]

Had a look at CircleCI, and some examples were unchanged (good), and this one looks better (good):

• https://output.circle-artifacts.com/output/job/bbac2ea9-489c-4f48-9442-2ba016761c66/artifacts/0/html/auto_examples/decoding/decoding_xdawn_eeg.html#sphx-glr-auto-examples-decoding-decoding-xdawn-eeg-py

Just a tiny minor things then I think we're good to go!

[Genuster]

@larsoner, if you don't have additional comments - it's ready ☺️

[larsoner]

Thanks @Genuster !