datalab_regression #796

mglowacki100 · 2023-08-07T07:40:42Z

Based on discussion #774
Preleminary work to ensure that default task (classification) will work smoothly. Still to do: label.py for regression, test for Datalab regression, docstring updates.

Usage example

from cleanlab import Datalab
import numpy as np

# Generate some toy-examples
N = 200
X = np.random.rand(N, 3)
y = 1 + 2 * X[:,0] + 3 * X[:,1] + 4 * X[:,2] + 0.05*np.random.randn(N)

# Run defaults
lab = Datalab(data={"y": y}, label_name='y', task="regression")
lab.find_issues(features=X)

# Run with any scikit-learn compatible model
from sklearn.neighbors import KNeighborsRegressor
lab = Datalab(data={"y": y}, label_name='y', task="regression")
model = KNeighborsRegressor(n_neighbors=3)
lab.find_issues(features=X, issue_types={"label": {"clean_learning_kwargs": {"model": model}}})

Preleminary work to ensure that default task (classification) will work smoothly.

CLAassistant · 2023-08-07T07:40:50Z

All committers have signed the CLA.

Removed unused imports

elisno

Great start @mglowacki100!

Here are a few comments.

cleanlab/datalab/internal/issue_manager_factory.py

cleanlab/datalab/internal/report.py

cleanlab/datalab/internal/issue_finder.py

elisno · 2023-08-07T15:30:29Z

You may want to merge master into your branch before moving on with development.

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

cleanlab/datalab/datalab.py

cleanlab/datalab/internal/data_issues.py

cleanlab/datalab/internal/issue_manager/regression/label.py

Doc-string added Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

comment explaining workaround

function renaming to avoid conflict

into dl_regression

name collision fix

add test class for the IssueFinder when the task is regression

The register method and issue manager factories will be more more maintainable if every issue manage is registered to a task. For now, only label issue checks work for regression, as that is the only issue manger registered for regression.

Run through cases where it looks at the kwargs. Another case where it looks at the issue_types kwarg.

not within the scope of this PR

docstrings, whitespace,...

elisno

Wonderful work @mglowacki100!

I've made some adjustments to the registry and related functionality, as well as handled labels are numerical values (for regression).

mglowacki100 · 2023-11-21T07:30:47Z

Hi @elisno
Thank you for your kind words! I really appreciate your acknowledgment of the work.
I look forward to continuing to contribute to cleanlab.
Best regards,
Marek

datalab_regression

4f6df23

Preleminary work to ensure that default task (classification) will work smoothly.

mglowacki100 added 8 commits August 7, 2023 09:42

Update label.py

3d58d98

Removed unused imports

black format, issue_name for regression

1c9fa64

Update issue_manager_factory.py

8a0a291

Update issue_manager_factory.py

d120d57

Update issue_manager_factory.py

29e9a83

Update issue_manager_factory.py

d202498

Update issue_manager_factory.py

46fafc5

Update issue_manager_factory.py

c5af243

elisno reviewed Aug 7, 2023

View reviewed changes

mglowacki100 and others added 17 commits August 7, 2023 17:51

Update cleanlab/datalab/internal/report.py

323b5d0

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/issue_finder.py

d167899

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/issue_manager_factory.py

430db78

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/issue_manager_factory.py

0050294

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/issue_manager_factory.py

3b4bbf9

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/issue_manager_factory.py

7c8cb7e

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update issue_manager_factory.py

02e8aac

fixing required positional argument: 'task'

817ce2e

reverse changes and change tests

922fb2f

change in architecture - test

e832db8

Update datalab.py

9afb9bb

Update issue_finder.py

8e4044d

Update report.py

d0f102b

fixing imagelab

fa1834a

fixing some errors

1a98816

fixing tests

9a6b6fd

Fixing tests - missing args

10b6753

elisno reviewed Aug 7, 2023

View reviewed changes

cleanlab/datalab/datalab.py Outdated Show resolved Hide resolved

elisno reviewed Oct 9, 2023

View reviewed changes

mglowacki100 and others added 11 commits October 15, 2023 21:15

Update cleanlab/datalab/internal/data_issues.py

52d4ee2

Doc-string added Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/data_issues.py

5bb6d3d

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/issue_manager/regression/label.py

290d19e

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/data_issues.py

e91fbc9

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

Update cleanlab/datalab/internal/data_issues.py

7448142

Co-authored-by: Elías Snorrason <eliassno@gmail.com>

method chaining

3057e14

Update issue_finder.py

5401665

comment explaining workaround

Merge branch 'cleanlab:master' into dl_regression

5de3fbc

Update datalab.py

91359e9

function renaming to avoid conflict

Merge branch 'dl_regression' of https://github.com/mglowacki100/cleanlab

e5c12e3

into dl_regression

Update datalab.py

a27f8ea

name collision fix

jwmueller requested a review from elisno October 18, 2023 02:19

elisno added 12 commits November 14, 2023 15:27

inject task into IssueFinder

0b9867f

make different strategies for getting available issue types

5bbf46e

add test class for the IssueFinder when the task is regression

apply black formatter

9d1a8e4

avoid mapping labels column for regression in Datalab

cf364c6

Pass in features to LabelIssueManager for regression

3540d62

Use single REGISTRY

585d14a

The register method and issue manager factories will be more more maintainable if every issue manage is registered to a task. For now, only label issue checks work for regression, as that is the only issue manger registered for regression.

explicitly test get_available_issue_types for classification tasks

3a77bcd

Run through cases where it looks at the kwargs. Another case where it looks at the issue_types kwarg.

apply black formatter

a16633a

remove class_imbalance from defaults

38e0901

not within the scope of this PR

minor updates to misc files

4c6faae

docstrings, whitespace,...

Merge branch 'master' into pr/mglowacki100/796

b467478

address missing imports for type-checking

204651c

elisno approved these changes Nov 20, 2023

View reviewed changes

elisno merged commit a6d1319 into cleanlab:master Nov 20, 2023
20 checks passed

elisno mentioned this pull request Nov 20, 2023

Extend regression in Datalab to handle predictions as input #899

Closed

elisno mentioned this pull request Nov 22, 2023

Predictions for regression in Datalab #902

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

datalab_regression #796

datalab_regression #796

mglowacki100 commented Aug 7, 2023 •

edited by elisno

Loading

CLAassistant commented Aug 7, 2023 •

edited

Loading

elisno left a comment

elisno commented Aug 7, 2023

elisno left a comment

mglowacki100 commented Nov 21, 2023

datalab_regression #796

datalab_regression #796

Conversation

mglowacki100 commented Aug 7, 2023 • edited by elisno Loading

Usage example

CLAassistant commented Aug 7, 2023 • edited Loading

elisno left a comment

Choose a reason for hiding this comment

elisno commented Aug 7, 2023

elisno left a comment

Choose a reason for hiding this comment

mglowacki100 commented Nov 21, 2023

mglowacki100 commented Aug 7, 2023 •

edited by elisno

Loading

CLAassistant commented Aug 7, 2023 •

edited

Loading