ordboost 0.3.0__tar.gz → 0.3.0.dev0__tar.gz

{ordboost-0.3.0 → ordboost-0.3.0.dev0}/PKG-INFO

@@ -1,10 +1,11 @@
 Metadata-Version: 2.4
 Name: ordboost
-Version: 0.3.0
+Version: 0.3.0.dev0
 Summary: Non-parametric discrete ordinal gradient boosting with probabilistic prediction calibration.
 Author-email: Mathias Roesler <mathias.roesler@auckland.ac.nz>
 License: Apache-2.0
 Keywords: calibration,gradient-boosting,machine-learning,ordinal-regression,probabilistic-forecasting,scikit-learn
+Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Science/Research
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.10
@@ -61,9 +62,7 @@ By framing continuous regression as a dynamic ordinal binning problem, **ordboos
 * **Non-Parametric Probabilistic Output**: Obtains complete predictive distribution objects capable of extracting probability mass functions (PMF), cumulative distribution functions (CDF), percentiles (`ppf`), and dynamic prediction intervals.
 * **Flexible Continuous Target Mapping**: Maps continuous values to discrete target spaces using configurable binning strategies (`QuantileBinMapper`, `UniformBinMapper`, `EmpiricalMeanBinMapper`, `EmpiricalMedianBinMapper`, `ContinuousBinMapper`).
 * **Monotonic Ordinal Constraints**: Supports constrained ordinal boosting (e.g., isotonic constraints) across sequential boundaries.
-* **Probabilistic Evaluation**: Evaluates probabilistic predictions directly using CRPS (`crps_score`), quantile loss (`pinball_loss`), and their skill scores.
-* **Prediction Intervals Measures**: Measures the prediction interval coverage (`interval_coverage_rate`), sharpness (`sharpness`), and Winkler score (`winkler_score`).
-* **Calibration Diagnostics**: Computes marginal calibration (`marginal_calibration_curve`) and probabilistic calibration (`pit_diagnostics`).
+* **Built-in Probabilistic Evaluation**: Evaluates probabilistic predictions directly using CRPS (`crps_score`), quantile loss (`pinball_loss`), prediction interval coverage (`interval_coverage_rate`), and Winkler scores (`winkler_score`).
 
 ---
 
@@ -174,4 +173,4 @@ for i in range(len(X_test)):
 
 ## Acknowledgements
 
-This package was developed using Gemini 3.6 Thinking and Claude Sonnet 5. The code was reviewed and edited by humans.
+This package was developed using Gemini 3.6 Thinking. The code was reviewed and edited by humans.

{ordboost-0.3.0 → ordboost-0.3.0.dev0}/README.md

@@ -29,9 +29,7 @@ By framing continuous regression as a dynamic ordinal binning problem, **ordboos
 * **Non-Parametric Probabilistic Output**: Obtains complete predictive distribution objects capable of extracting probability mass functions (PMF), cumulative distribution functions (CDF), percentiles (`ppf`), and dynamic prediction intervals.
 * **Flexible Continuous Target Mapping**: Maps continuous values to discrete target spaces using configurable binning strategies (`QuantileBinMapper`, `UniformBinMapper`, `EmpiricalMeanBinMapper`, `EmpiricalMedianBinMapper`, `ContinuousBinMapper`).
 * **Monotonic Ordinal Constraints**: Supports constrained ordinal boosting (e.g., isotonic constraints) across sequential boundaries.
-* **Probabilistic Evaluation**: Evaluates probabilistic predictions directly using CRPS (`crps_score`), quantile loss (`pinball_loss`), and their skill scores.
-* **Prediction Intervals Measures**: Measures the prediction interval coverage (`interval_coverage_rate`), sharpness (`sharpness`), and Winkler score (`winkler_score`).
-* **Calibration Diagnostics**: Computes marginal calibration (`marginal_calibration_curve`) and probabilistic calibration (`pit_diagnostics`).
+* **Built-in Probabilistic Evaluation**: Evaluates probabilistic predictions directly using CRPS (`crps_score`), quantile loss (`pinball_loss`), prediction interval coverage (`interval_coverage_rate`), and Winkler scores (`winkler_score`).
 
 ---
 
@@ -142,4 +140,4 @@ for i in range(len(X_test)):
 
 ## Acknowledgements
 
-This package was developed using Gemini 3.6 Thinking and Claude Sonnet 5. The code was reviewed and edited by humans.
+This package was developed using Gemini 3.6 Thinking. The code was reviewed and edited by humans.

{ordboost-0.3.0 → ordboost-0.3.0.dev0}/pyproject.toml

@@ -8,6 +8,7 @@ authors = [{ name = "Mathias Roesler", email = "mathias.roesler@auckland.ac.nz"
 dynamic = ["version"]
 keywords = ["calibration", "gradient-boosting", "machine-learning", "ordinal-regression", "probabilistic-forecasting", "scikit-learn"]
 classifiers = [
+  "Development Status :: 3 - Alpha",
   "Intended Audience :: Science/Research",
   "Programming Language :: Python :: 3",
   "Programming Language :: Python :: 3.10",

{ordboost-0.3.0 → ordboost-0.3.0.dev0}/src/ordboost/__init__.py

@@ -14,21 +14,14 @@ from ordboost.mappers import (
     UniformBinMapper,
 )
 from ordboost.metrics import (
-    baseline_distribution,
     crps_score,
-    crps_skill_score,
     interval_coverage_rate,
-    marginal_calibration_curve,
     pinball_loss,
-    pinball_loss_skill_score,
-    pit_diagnostics,
-    pit_ks_test,
-    sharpness,
     winkler_score,
 )
 from ordboost.models import OrdBoostClassifier, OrdBoostRegressor
 
-__version__ = "0.3.0"
+__version__ = "0.3.0.dev0"
 
 __all__ = [
     # Models
@@ -46,15 +39,8 @@ __all__ = [
     "DiscretePredictiveDistribution",
     "ContinuousPredictiveDistribution",
     # Metrics
-    "baseline_distribution",
     "crps_score",
-    "crps_skill_score",
     "interval_coverage_rate",
-    "marginal_calibration_curve",
-    "sharpness",
     "pinball_loss",
-    "pinball_loss_skill_score",
-    "pit_diagnostics",
-    "pit_ks_test",
     "winkler_score",
 ]

{ordboost-0.3.0 → ordboost-0.3.0.dev0}/src/ordboost/models.py

@@ -454,7 +454,7 @@ class OrdBoostRegressor(BaseEstimator, RegressorMixin):
     bin_strategy : {"quantile", "uniform"}, default="quantile"
         Strategy used to define automatic bin boundaries when `bin_edges`
         is None.
-    mapper : BaseBinMapper, {"median", "mean", "quantile", "uniform", "continuous"} or None, default="median"
+    mapper : {"median", "mean", "quantile", "uniform", "continuous"}, BaseBinMapper, or None, default="median"
         Bin mapping strategy used to convert predicted PMFs back to
         continuous predictions. A string selects the corresponding
         `BaseBinMapper` subclass, constructed automatically with

{ordboost-0.3.0 → ordboost-0.3.0.dev0}/src/ordboost.egg-info/PKG-INFO

@@ -1,10 +1,11 @@
 Metadata-Version: 2.4
 Name: ordboost
-Version: 0.3.0
+Version: 0.3.0.dev0
 Summary: Non-parametric discrete ordinal gradient boosting with probabilistic prediction calibration.
 Author-email: Mathias Roesler <mathias.roesler@auckland.ac.nz>
 License: Apache-2.0
 Keywords: calibration,gradient-boosting,machine-learning,ordinal-regression,probabilistic-forecasting,scikit-learn
+Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Science/Research
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.10
@@ -61,9 +62,7 @@ By framing continuous regression as a dynamic ordinal binning problem, **ordboos
 * **Non-Parametric Probabilistic Output**: Obtains complete predictive distribution objects capable of extracting probability mass functions (PMF), cumulative distribution functions (CDF), percentiles (`ppf`), and dynamic prediction intervals.
 * **Flexible Continuous Target Mapping**: Maps continuous values to discrete target spaces using configurable binning strategies (`QuantileBinMapper`, `UniformBinMapper`, `EmpiricalMeanBinMapper`, `EmpiricalMedianBinMapper`, `ContinuousBinMapper`).
 * **Monotonic Ordinal Constraints**: Supports constrained ordinal boosting (e.g., isotonic constraints) across sequential boundaries.
-* **Probabilistic Evaluation**: Evaluates probabilistic predictions directly using CRPS (`crps_score`), quantile loss (`pinball_loss`), and their skill scores.
-* **Prediction Intervals Measures**: Measures the prediction interval coverage (`interval_coverage_rate`), sharpness (`sharpness`), and Winkler score (`winkler_score`).
-* **Calibration Diagnostics**: Computes marginal calibration (`marginal_calibration_curve`) and probabilistic calibration (`pit_diagnostics`).
+* **Built-in Probabilistic Evaluation**: Evaluates probabilistic predictions directly using CRPS (`crps_score`), quantile loss (`pinball_loss`), prediction interval coverage (`interval_coverage_rate`), and Winkler scores (`winkler_score`).
 
 ---
 
@@ -174,4 +173,4 @@ for i in range(len(X_test)):
 
 ## Acknowledgements
 
-This package was developed using Gemini 3.6 Thinking and Claude Sonnet 5. The code was reviewed and edited by humans.
+This package was developed using Gemini 3.6 Thinking. The code was reviewed and edited by humans.