Skore 0.8.2

⚠️ change in feature_importance API.

New feature

• Add timing metrics in reports, in #1454 and #1481

Improvement

• Renamed report.feature_importance.feature_permutation to report.feature_importance.permutation in #1465
• Change the default of aggregate parameter for cross-validation report, in #1440
• Add get_predictions method to all reporters, in #1489
• Make error in __init__ more copy-able, in #1447
• Implement __repr__ for project objects, in #1448
• Make EstimatorReport.compute_metric_scores return metrics in native types, in #1462
• Add optional argument to return dict in train_test_split, in #1460
• Use a more general formulation for the rich repr for report objects, in #1499
• Rename scatter_kwargs and line_kwargs in PredictionErrorDisplay, in #1474
• Improve plots aspect, in #1485

Bugfix

• Do not expose Brier score if estimator does not implement predict_proba in #1471
• Make sure that average != None return a float for metric in #1502
• Check if private metrics are available for an estimator in #1496
• Support joblib < 1.4 by using list container as output in #1492
• Avoid pandas warning by using float('nan') instead of None in #1484

---

Full Changelog: 0.8.1...0.8.2

Skore 0.8.1

In this release, only updates to the documentation.

---

Full Changelog: 0.8.0...0.8.1

Skore 0.8.0

New features

The EstimatorReport is enhanced and enriched with feature importance! It can be computed in three different ways: feature permutation, model coefficient for linear models, and mean impurity decrease for tree-based models.

• Compute the model coefficients for linear models, in #1339
• Compute the feature permutation importance and mean decrease impurity, in #1365

Improments

• Add a method .set_style to displays, in #1336
• Give a description to attributes in helpers, in #1366

Bugfix

• Accept Target expressed in Series again, in #1327

Removed

Removed the Skore-lib UI.

---

Full Changelog: 0.7.1...0.8.0

Skore 0.7.1

New features

• feat: Add dedicated Display classes to ComparisonReport, in #1309
• feat: Add cache_predictions method to ComparisonReport, in #1352

Improvements

• feat(api): Allow to flatten index in reports, in #1300

---

Full Changelog: 0.7.0...0.7.1

Skore 0.7.0

This release contains breaking changes! The CrossValidationReporter was removed.

New features

• Add ComparisonReport to compare instances of EstimatorReport, in #1286
• Add config manager to manage progress bar, in #1334

Improvements

• Remove arrows in the metric names in dataframe to ease its use, in #1329
• Return most native type for metrics in EstimatorReport, in #1283
• Improve the API of EstimatorReport for plots, in #1297 and #1299
• Improve how a project is created and deleted, in #1194 and #1322

Bugfixes

• Fix the residual plot which may be displayed non properly, in #1335

Removed

• Remove CrossValidationReporter, in #1265

---

Full Changelog: 0.6.1...0.7.0

Skore 0.6.1

New features

Nothing in this edition. Keep posted to see what's next!

Improvements

Many documentation improvements.

Bugfixes

Nothing necessary for now!

---

Full Changelog: 0.6.0...0.6.1

Skore 0.6.0

This release adds several major features. It contains breaking changes.

New features

• Add EstimatorReport in #997
• Revamp CrossValidationReport thanks to EstimatorReport in #1091
• Add notes to the version of an item in #1084, #1101, #1138
• Enable what you put is what you get in #1052

Features improvement

• Prettify warning in train_test_split in #1086
• Refine ML task detection logic in #1133
• Add markdown editor to annotate item in #1082
• Improve the format of numbers in cv report in #1148

Bugfix

• several UI improvements

---

Full Changelog: 0.5.1...0.6.0

Skore 0.6 - rc 1

This release adds several major features. It contains breaking changes.

New features

• Add EstimatorReport in #997
• Revamp CrossValidationReport thanks to EstimatorReport in #1091
• Add notes to the version of an item in #1084, #1101, #1138
• Enable what you put is what you get in #1052

Features improvement

• Prettify warning in train_test_split in #1086
• Refine ML task detection logic in #1133
• Add markdown editor to annotate item in #1082
• Improve the format of numbers in cv report in #1148

Bugfix

• several UI improvements

---

Full Changelog: 0.5.1...0.6.0-rc.0

Skore 0.5.1

New features

• Single open command instead of create and load commands in #992
• Light/dark theme toggle in skore UI in #988

Improvements

• Replace neg_* metrics with their positive counterparts in #975
• Split time-related results of CrossValidationReporter into their own plots in #986
• Improve colors and add friendly verbose in CLI in #960 & #962
• Use skore-ui widgets in notebooks in #926
• Indicate whether metrics are lower/higher-means-better in CrossValidationReporter in #1053

Bugfixes

• fix: Fix quickstart instruction in CLI in #1048
• fix: Make skore.show_versions() robust to version specifiers in requirements in #1056

---

Full Changelog: 0.5.0...0.5.1

Skore 0.5.0

New features

• ActivityFeed
  • Conveniently view all items as they are created.
  • PRs: #857 and #848
• CrossValidationReporter
  • Modular cross validation suite, our goal is to make it comprehensive.
  • PRs: #847 and #871

Improvements

• Add fit_time_per_data_point and score_time_per_data_point metrics in CrossValidationReporter plot in #875
• Many miscellaneous improvements

Bugfixes

• Skore UI now updates the connection to the project when it's overwritten in #892
• Warning about undefined pandas dependency removed in #867

Removed

• Remove CrossValidationAggregationItem in #922

---

Full Changelog: https://github.com/probabl-ai/skore/commits/0.5.0