Open fkiraly opened 5 months ago
PS: we weren't sure about authorship or maintainership - kindly be welcome to add yourself/yourselves to the author or maintainer fields of any of the adapters. Maintainer field only if you would like to take on maintenance of the adapter in sktime
, of course. Both fields are shown in the searchable estimator overview and are one of our primary means of giving individual credit: https://www.sktime.net/en/latest/estimator_overview.html
I just went ahead and added you to the author fields, @johannfaouzi: https://github.com/sktime/sktime/pull/6270 You will be visible in the estimator overview in first position.
Let us know of any comments or questions.
While I'm here, I wanted to point out that at
sktime
, we've started to interfacepyts
algorithms, upon popular demand, to have them indexed for users who are searching for TSC and time series transformations.Users will be able to find interfaced estimators using the indexing utility
all_estimators
insktime.registry
, and use them as components insktime
pipelines and composites. For this, users will need to installpyts
, and they recieve an informative error message to this effect when they attempt to construct pipelines. Proper credit topyts
is of course also given, by name of estimator (usespyts
brand) and in docstring, feedback appreciated.The estimators are also regularly tested against standard API contracts, that's how we found #158.
If you would like to help out, or observe:
sktime
is here, feel free to comment: https://github.com/sktime/sktime/issues/5850pyts
estimators is formulaic, we have written a general adapter that can be inherited fromWe're not sure yet about distances.
Further, the knn classifier is neat, but we're wondering what is the best way to allow it to take abstract distances, which are first order citizens in
sktime
(also estimators). Perhaps there is a collaboration opportunity here.