Franz Király
Franz Király
Regarding unsupervised models such as PCA, kmeans, etc discussed in #44. I know these are commonly encapsulated within the transformer formalism, but it would do the methodology behind them injustice...
This is somewhat related to #85 (visualizations for tuning) but I'm opening an issue as it is not the same. A crucial part of the modelling workflow are diagnostics of...
Small changes to improve the documentation by pointing towards the example notebooks, removing outdated notebooks from the sphinx docs. See towards https://github.com/alan-turing-institute/sktime/issues/2127. Changes: * "tutorials" is now a list of...
Update: not only do the PR retrieved fluctuate in/out, it seems that the changelog generator missed a substantial fraction of changes - around 40 PR and 7 contributors between 0.13.0...
The test module`transformations.series.tests.test_date` fits some estimators directly in the module, this would fail at test collection time rather than at test time (where it should). The fitting should be moved...
This PR robustifies `BaseObject.set_tags` against forgotten `__init__`, e.g., if a child object forgets to `super.__init__` (or intentionally does not call `super.__init__`). Without this, an unexpected exception can occur if a...
I think we need a small design document for the rework of the reducers. FYI @danbartl, given recent discussions. Cases we should cover with modular design: * lagging `X` and...
We have now implemented some experimental algorithms, and collected observations on common inputs and outputs: https://github.com/sktime/sktime/discussions/2848 We should try to consolidate this into ideas for a base class design (strategy...
DO NOT MERGE, DO NOT CLOSE PERMANENT DIAGNOSTIC PR to investigate whether soft dependencies have been correctly isolated in tests. Based on PRs: * PR https://github.com/alan-turing-institute/sktime/pull/2985 which ensures the main...
See here for overview of recent test framework related PRs: https://github.com/alan-turing-institute/sktime/issues/2053 --- Experimental version of #1992 with: * generators for pytest fixture combinations * hash-caching of fitted estimators to avoid...