Add files via upload

Author: andyjiayuwang

pyproject.toml

@@ -0,0 +1,20 @@
+[build-system]
+requires = ["setuptools>=68", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "rlearner"
+version = "0.1.0"
+description = "Python tools for R-learner nuisance estimation and CATE workflows."
+requires-python = ">=3.10"
+dependencies = [
+    "numpy>=1.24",
+    "scikit-learn>=1.3",
+    "scipy>=1.11",
+]
+
+[tool.setuptools]
+package-dir = {"" = "src"}
+
+[tool.setuptools.packages.find]
+where = ["src"]

src/rlearner/__init__.py

@@ -0,0 +1,42 @@
+"""Public package exports for the initial R-learner package."""
+
+from .evaluation import (
+    BLPTestResult,
+    CalibrationBinResult,
+    CalibrationTestResult,
+    CoefficientInference,
+    UpliftCurvePoint,
+    UpliftTestResult,
+    blp_test,
+    calibration_test,
+    uplift_test,
+)
+from .learner import RLearner
+from .nuisance import (
+    CrossFittedNuisanceEstimator,
+    ManualNuisanceEstimator,
+    NuisanceResult,
+    SuperLearnerClassifier,
+    SuperLearnerRegressor,
+)
+from .r_loss import RLossStacking, RLossWrapper
+
+__all__ = [
+    "BLPTestResult",
+    "CalibrationBinResult",
+    "CalibrationTestResult",
+    "CoefficientInference",
+    "CrossFittedNuisanceEstimator",
+    "ManualNuisanceEstimator",
+    "NuisanceResult",
+    "RLearner",
+    "RLossStacking",
+    "RLossWrapper",
+    "SuperLearnerClassifier",
+    "SuperLearnerRegressor",
+    "UpliftCurvePoint",
+    "UpliftTestResult",
+    "blp_test",
+    "calibration_test",
+    "uplift_test",
+]

src/rlearner/__pycache__/__init__.cpython-312.pyc

@@ -0,0 +1,62 @@
+"""Validation helpers shared by nuisance-estimation components."""
+
+from __future__ import annotations
+
+from typing import Any
+
+import numpy as np
+from sklearn.utils.validation import check_array
+
+
+def validate_features(X: Any) -> np.ndarray:
+    """Validate and coerce a feature matrix to a dense 2D numpy array."""
+    return check_array(X, ensure_2d=True, dtype=None, ensure_all_finite="allow-nan")
+
+
+def validate_vector(values: Any, *, name: str, n_samples: int | None = None) -> np.ndarray:
+    """Validate and coerce a one-dimensional target or prediction vector."""
+    vector = check_array(values, ensure_2d=False, dtype=None, ensure_all_finite="allow-nan")
+    vector = np.asarray(vector).reshape(-1)
+
+    if n_samples is not None and vector.shape[0] != n_samples:
+        raise ValueError(f"{name} must have length {n_samples}, got {vector.shape[0]}.")
+
+    return vector
+
+
+def validate_binary_treatment(d: Any, *, n_samples: int | None = None) -> np.ndarray:
+    """Validate that treatment assignments are binary-coded."""
+    treatment = validate_vector(d, name="d", n_samples=n_samples)
+    unique_values = np.unique(treatment)
+
+    if unique_values.size > 2 or not np.all(np.isin(unique_values, [0, 1])):
+        raise ValueError(
+            "Binary treatment is required for the built-in nuisance estimators. "
+            "Expected values in {0, 1}."
+        )
+
+    return treatment.astype(int, copy=False)
+
+
+def validate_manual_predictions(
+    *,
+    y: Any,
+    d: Any,
+    y_hat: Any,
+    d_hat: Any,
+    propensity_clip: float,
+) -> tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
+    """Validate manual nuisance predictions against observed outcomes and treatment."""
+    y_array = validate_vector(y, name="y")
+    d_array = validate_binary_treatment(d, n_samples=y_array.shape[0])
+    y_hat_array = validate_vector(y_hat, name="y_hat", n_samples=y_array.shape[0])
+    d_hat_array = validate_vector(d_hat, name="d_hat", n_samples=y_array.shape[0])
+
+    if np.any(~np.isfinite(y_hat_array)):
+        raise ValueError("y_hat must contain only finite values.")
+    if np.any(~np.isfinite(d_hat_array)):
+        raise ValueError("d_hat must contain only finite values.")
+
+    d_hat_array = np.clip(d_hat_array.astype(float, copy=False), propensity_clip, 1.0 - propensity_clip)
+
+    return y_array, d_array, y_hat_array, d_hat_array