feat: added halfnormal distribution (#387)

Author: acdundore

ngboost/distns/__init__.py

@@ -4,6 +4,7 @@
 from .distn import ClassificationDistn, Distn, RegressionDistn
 from .exponential import Exponential
 from .gamma import Gamma
+from .halfnormal import HalfNormal
 from .laplace import Laplace
 from .lognormal import LogNormal
 from .multivariate_normal import MultivariateNormal
@@ -21,6 +22,7 @@
     "RegressionDistn",
     "Exponential",
     "Gamma",
+    "HalfNormal",
     "Laplace",
     "LogNormal",
     "MultivariateNormal",

ngboost/distns/halfnormal.py

@@ -0,0 +1,55 @@
+"""The NGBoost Half-Normal distribution and scores"""
+import numpy as np
+from scipy.stats import halfnorm as dist
+
+from ngboost.distns.distn import RegressionDistn
+from ngboost.scores import LogScore
+
+
+class HalfNormalLogScore(LogScore):
+    def score(self, Y):
+        return -self.dist.logpdf(Y)
+
+    def d_score(self, Y):
+        D = np.zeros((len(Y), 1))
+        D[:, 0] = (self.scale**2 - Y**2) / self.scale**2  # dL/d(log(scale))
+        return D
+
+    def metric(self):
+        FI = 2 * np.ones_like(self.scale)
+        return FI[:, np.newaxis, np.newaxis]
+
+
+class HalfNormal(RegressionDistn):
+    """
+    Implements the Half-Normal distribution for NGBoost.
+
+    The Half-Normal distribution has one parameter, scale.
+    The scipy loc parameter is held constant at zero for this implementation.
+    LogScore is supported for the Half-Normal distribution.
+    """
+
+    n_params = 1
+    scores = [HalfNormalLogScore]
+
+    # pylint: disable=super-init-not-called
+    def __init__(self, params):
+        self._params = params
+        self.scale = np.exp(params[0])  # scale (sigma)
+        self.dist = dist(loc=0, scale=self.scale)
+
+    def fit(Y):
+        _loc, scale = dist.fit(Y, floc=0)  # loc held constant
+        return np.array([np.log(scale)])
+
+    def sample(self, m):
+        return np.array([self.dist.rvs() for i in range(m)])
+
+    def __getattr__(self, name):
+        if name in dir(self.dist):
+            return getattr(self.dist, name)
+        return None
+
+    @property
+    def params(self):
+        return {"loc": np.zeros(shape=self.scale.shape), "scale": self.scale}

tests/test_distns.py

@@ -13,6 +13,7 @@
     Distn,
     Exponential,
     Gamma,
+    HalfNormal,
     LogNormal,
     MultivariateNormal,
     Normal,
@@ -68,6 +69,7 @@ def is_t_distribution(
         Normal,
         NormalFixedVar,
         NormalFixedMean,
+        HalfNormal,
         LogNormal,
         Exponential,
         Gamma,

tests/test_score.py

@@ -9,6 +9,7 @@
     Cauchy,
     Distn,
     Gamma,
+    HalfNormal,
     Laplace,
     MultivariateNormal,
     Normal,
@@ -97,6 +98,7 @@ def idfn(dist_score: DistScore):
 TEST_METRIC: List[DistScore] = [
     (Normal, LogScore),
     (Normal, CRPScore),
+    (HalfNormal, LogScore),
     (TFixedDfFixedVar, LogScore),
     (Laplace, LogScore),
     (Poisson, LogScore),