[ENH] Add online regressor support and River adapter by patelchaitany · Pull Request #1068 · sktime/skpro

patelchaitany · 2026-06-10T19:29:12Z

Reference Issues/PRs

What does this implement/fix? Explain your changes.

This PR adds the foundation for online regression in skpro, including a new regressor_online scitype and BaseOnlineRegressor with capability:update and capability:pred_int. It introduces the capability:pred_int tag and gates probabilistic prediction methods on regressors that do not support them. It also adds RiverRegressor as a point-prediction adapter for River models, with deepcopy-based cloning and coerce_to_skpro_regressor() for wrapping raw River estimators.

Does your contribution introduce a new dependency? If yes, which one?

No

What should a reviewer concentrate their feedback on?

Did you add any tests for the change?

Any other comments?

PR checklist

For all contributions

I've added myself to the list of contributors with any new badges I've earned :-)
How to: add yourself to the all-contributors file in the skpro root directory (not the CONTRIBUTORS.md). Common badges: code - fixing a bug, or adding code logic. doc - writing or improving documentation or docstrings. bug - reporting or diagnosing a bug (get this plus code if you also fixed the bug in the PR).maintenance - CI, test framework, release.
See here for full badge reference
The PR title starts with either [ENH], [MNT], [DOC], or [BUG]. [BUG] - bugfix, [MNT] - CI, test framework, [ENH] - adding or improving code, [DOC] - writing or improving documentation or docstrings.

For new estimators

I've added the estimator to the API reference - in docs/source/api_reference/taskname.rst, follow the pattern.
I've added one or more illustrative usage examples to the docstring, in a pydocstyle compliant Examples section.
If the estimator relies on a soft dependency, I've set the python_dependencies tag and ensured
dependency isolation, see the estimator dependencies guide.

fkiraly

Nice! Though I think there is somet duplication going on.

I think the new base class, BaseOnlineRegressor, can be merged into BaseProbaRegressor, there is no substantial added functionality.
The new jackknife-plus regressor can be merged into the existing one (the non-MAPIE regressor), simply add the update logic.

felipeangelimvieira · 2026-06-18T12:03:32Z

+        -------
+        self : reference to self
+        """
+        errs, bs_vs_ix_new, estimators = self._bootstrap_batch(X, y, self.estimators_)


As I understand, update is behaving just like a brand new fit on the new batch. Shouldn't _boostrap_batch call _update on the new estimators instead of fit, if the user hass called EnbpiRegressor.update?

I think I made mistake The original EnbpiRegressor dose not update the trained model it only update the residuals

felipeangelimvieira · 2026-07-01T23:21:08Z

        return TestAllRegressors


+class regressor_online(_BaseScitypeOfObject):


Now that there is no BaseOnlineRegressor, we can remove this object type.

Yes we can remove that but for the river using we are using separate tag named regressor_online and in future if create separate test for online_regressor it would be easy to add test, wdyt.

Our API won't be supporting online regression directly, right? Except for update. I believe it would better to follow YAGNI than add this tag at the moment. In addition, this would need to be a separate tag, not a object type value.

felipeangelimvieira · 2026-07-01T23:22:01Z

    }


+class capability__pred_int(_BaseTag):


Could you explain the motivation for this new tag?

Until now all skpro regressors supported probabilistic prediction, so this tag wasn't needed. The River adapter is the first point-prediction-only estimator, so capability:pred_int is needed to distinguish which regressors support predict_interval/predict_proba - matching sktime's convention.

Add River adapter package (utils, clone, RiverRegressor) and a coerce_to_skpro_regressor helper. Introduce BaseOnlineRegressor and capability:pred_int tag; update BaseProbaRegressor to check pred_int for probabilistic methods.

fkiraly · 2026-08-06T13:48:31Z

+    bool
+        True if ``obj`` is a River estimator, False otherwise.
+    """
+    if obj is None:


I think you can preface this with a _check_soft_dependencies call, then you can assume that river is present. That is: if check_soft_dependencies("river"...) is False, return False. Anything that comes after can assume that river is present in the environment.

Why this works: if river is not present, the passed object cannot be a river instance.

fkiraly · 2026-08-06T13:50:19Z

@@ -294,6 +298,7 @@ def predict_proba(self, X):
        y : skpro BaseDistribution, same length as `X`
            labels predicted for `X`
        """


are we missing a check_is_fitted here? Unrelated to your PR but it is still missing and shoudl be fixed?

fkiraly · 2026-08-06T13:52:41Z

    "matplotlib>=3.3.2",
    "polars<1.43.0",
    "pymc; python_version < '3.13'",
+    "river",


we should not add it here - this should be an estimator specific dependency.

fkiraly

Great, thanks!

Since this would now add "online-only" regressors to the zoo, we also need to check compatibility with:

tuners, grid search, etc
pipelines
further compositions, if any

For instance, there also, the capability:pred_int tag needs to be set consistently with a potential online update regressor.

You could add this in this PR together with tests, or in a PR building on top of this.

Implement online EnbPI: do not refit clones on update; predict with stored estimators, append new residuals and slide out oldest entries to maintain a fixed pool size. Refactor _bootstrap_batch to initialize estimators_ and fix prediction/residual array shapes and indexing. Signed-off-by: Chaitany Patel <patelchaitany93@gmail.com>

Signed-off-by: Chaitany Patel <patelchaitany93@gmail.com>

…gressor Signed-off-by: Chaitany Patel <patelchaitany93@gmail.com> # Conflicts: # skpro/regression/online/_dont_refit.py # skpro/regression/online/_refit.py # skpro/regression/online/_refit_every.py

patelchaitany requested review from felipeangelimvieira and fkiraly as code owners June 10, 2026 19:29

patelchaitany force-pushed the enh/river-online-regressor branch from 88ea31c to 732a597 Compare June 10, 2026 19:31

patelchaitany marked this pull request as draft June 10, 2026 19:32

patelchaitany force-pushed the enh/river-online-regressor branch 2 times, most recently from 6ca8dfc to 57f3208 Compare June 11, 2026 06:41

patelchaitany marked this pull request as ready for review June 11, 2026 09:17

fkiraly mentioned this pull request Jun 13, 2026

[ENH] update capability for GridSearchCV and RandomizedSearchCV #1050

Open

patelchaitany force-pushed the enh/river-online-regressor branch 2 times, most recently from ed314f0 to 4424926 Compare June 14, 2026 02:13

fkiraly requested changes Jun 14, 2026

View reviewed changes

patelchaitany force-pushed the enh/river-online-regressor branch from 4424926 to 90cd8a6 Compare June 15, 2026 08:05

patelchaitany requested a review from fkiraly June 15, 2026 08:09

fkiraly assigned patelchaitany Jun 16, 2026

felipeangelimvieira reviewed Jun 18, 2026

View reviewed changes

patelchaitany requested a review from felipeangelimvieira June 18, 2026 16:22

patelchaitany force-pushed the enh/river-online-regressor branch from d64477d to 43fd983 Compare June 18, 2026 16:32

felipeangelimvieira requested changes Jul 1, 2026

View reviewed changes

patelchaitany added 2 commits July 2, 2026 11:51

Add River online regressor adapter and base class

bbbc1f4

Add River adapter package (utils, clone, RiverRegressor) and a coerce_to_skpro_regressor helper. Introduce BaseOnlineRegressor and capability:pred_int tag; update BaseProbaRegressor to check pred_int for probabilistic methods.

Add ConformalResidualRegressor

5b9a862

patelchaitany force-pushed the enh/river-online-regressor branch from 43fd983 to bde44e9 Compare July 2, 2026 06:21

patelchaitany requested a review from felipeangelimvieira July 6, 2026 07:30

patelchaitany force-pushed the enh/river-online-regressor branch 2 times, most recently from cfe3cfb to 871c953 Compare July 14, 2026 15:57

felipeangelimvieira approved these changes Jul 16, 2026

View reviewed changes

fkiraly reviewed Aug 6, 2026

View reviewed changes

fkiraly requested changes Aug 6, 2026

View reviewed changes

patelchaitany force-pushed the enh/river-online-regressor branch from 871c953 to 8c679cd Compare August 10, 2026 13:39

patelchaitany requested a review from fkiraly August 10, 2026 13:40

patelchaitany added 2 commits August 12, 2026 00:59

Set capability:pred_int consistently across compositions

723f20b

Signed-off-by: Chaitany Patel <patelchaitany93@gmail.com>

Merge remote-tracking branch 'upstream/main' into enh/river-online-re…

4346f4c

…gressor Signed-off-by: Chaitany Patel <patelchaitany93@gmail.com> # Conflicts: # skpro/regression/online/_dont_refit.py # skpro/regression/online/_refit.py # skpro/regression/online/_refit_every.py

Navigation Menu

Uh oh!

[ENH] Add online regressor support and River adapter - #1068

[ENH] Add online regressor support and River adapter#1068
patelchaitany wants to merge 5 commits into
sktime:mainfrom
patelchaitany:enh/river-online-regressor

patelchaitany commented Jun 10, 2026

Uh oh!

fkiraly left a comment

Uh oh!

felipeangelimvieira Jun 18, 2026

Uh oh!

patelchaitany Jun 18, 2026

Uh oh!

felipeangelimvieira Jul 1, 2026

Uh oh!

patelchaitany Jul 2, 2026

Uh oh!

felipeangelimvieira Jul 2, 2026

Uh oh!

felipeangelimvieira Jul 1, 2026

Uh oh!

patelchaitany Jul 2, 2026

Uh oh!

fkiraly Aug 6, 2026

Uh oh!

fkiraly Aug 6, 2026

Uh oh!

fkiraly Aug 6, 2026

Uh oh!

fkiraly left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		return TestAllRegressors


		class regressor_online(_BaseScitypeOfObject):

Uh oh!

Conversation

patelchaitany commented Jun 10, 2026

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Does your contribution introduce a new dependency? If yes, which one?

What should a reviewer concentrate their feedback on?

Did you add any tests for the change?

Any other comments?

PR checklist

For all contributions

For new estimators

Uh oh!

fkiraly left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fkiraly left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants