Summary of the Capabilities of time series classifiers

summary of the type of data input the main sktime classifiers can deal with. If you are interested in extending the capabilities of any of these classifiers, get in touch and we will offer some guidance.

FK comments [added 2021-07-21]: before implementing this, the part on _tags in #1146 should be done, i.e., moving the capabilities to the new _tags system compatible with BaseObject. This should be done in collaboration with @TonyBagnall and either myself or @mloning, since it touches some design specifications that we need to follow.

There is also the automated estimator overview in the sphinx docs which can be taken as a template for this, see #704.

[fkiraly]

@fkiraly If I am not wrong here, is this kinda extension to Automated learning algorithm overview?

Correct, on the docs side.

Under the hood, we also need a consolidated tags system, as in #981, with the registry retrieval/query functionality that @mloning proposed.

[TonyBagnall]

adding to this is to add Cababilities (soon to be tags) to classifiers without them, including [ColumnEnsembleClassifier, ComposableTimeSeriesForestClassifier, IndividualBOSS, IndividualTDE, ProximityStump, ProximityTree]