Merge branch 'develop' into release-1.10

Author: simomarsili

tests/inputs.py

@@ -0,0 +1,115 @@
+# -*- coding: utf-8 -*-
+# pylint: disable=missing-docstring
+"""Tests for the entropy() function."""
+import numpy
+
+from make_test_ref import SEED
+
+
+class Pmf:
+    """PMF class.
+
+    Parameters
+    ----------
+    alpha : float
+        Concentration parameter.
+    k : int
+        Alphabet size.
+    zero : float or None
+        Fraction of bins with exactly zero probability.
+
+    """
+
+    def __init__(self, alpha=0.1, k=10000, zero=0):
+        numpy.random.seed(SEED)
+        self.alpha = alpha
+        self.k = k
+        self.zero = zero
+        self._pk = self._generate_pk(self.alpha, self.k, self.zero)
+        self._entropy = None
+
+    @property
+    def pk(self):
+        return self._pk
+
+    @staticmethod
+    def _generate_pk(alpha, k, zero=0):
+        """Return a Dirichlet sample."""
+        pk = numpy.random.dirichlet([alpha] * k)
+        if zero:
+            n_zero = numpy.random.binomial(k, zero)
+            pk[:n_zero] = 0
+            pk /= pk.sum()
+            pk = pk[n_zero:]
+        return pk
+
+    def randomize(self):
+        """Reset pk to a random pmf."""
+        self._pk = self._generate_pk(self.alpha, self.k, self.zero)
+        self._entropy = None
+        return self
+
+    @staticmethod
+    def entropy_from_pmf(a):
+        pk = numpy.asarray(a)
+        pk = pk[pk > 0]
+        return -numpy.sum(pk * numpy.log(pk))
+
+    @property
+    def entropy(self):
+        """Entropy for PMF"""
+        if self._entropy is None:
+            self._entropy = self.entropy_from_pmf(self.pk)
+        return self._entropy
+
+
+class Counts:
+    def __init__(self, n=100, pmf=None, **kwargs):
+        """
+        Counts class.
+
+        Parameters
+        ----------
+        n : int
+            Number of samples.
+        pmf : Pmf object, optional
+            Alphabet size.
+
+        """
+        numpy.random.seed(SEED)
+        self.n = n
+
+        if pmf is not None:
+            self.pmf = pmf
+        else:
+            self.pmf = Pmf(**kwargs)
+
+        self._nk = self._generate_nk(self.n, self.pmf.pk)
+        self._entropy = None
+
+    @property
+    def nk(self):
+        return self._nk
+
+    @staticmethod
+    def _generate_nk(n, pk):
+        """Return a Multinomial sample."""
+        return numpy.random.multinomial(n, pk)
+
+    def randomize(self):
+        self.nk = self._generate_nk(self.n, self.pmf.pk)
+        self._entropy = None
+        return self
+
+    @staticmethod
+    def entropy_from_counts(a, estimator, **kwargs):
+        nk = numpy.asarray(a)
+        estimator.fit(nk, **kwargs)
+        return estimator.estimate_
+
+    def entropy(self, estimator, **kwargs):
+        """Entropy estimate from counts using `estimator`."""
+        if self._entropy is None:
+            self._entropy = self.entropy_from_counts(self.nk, estimator,
+                                                     **kwargs)
+        return self._entropy

tests/test_entropy.py

@@ -0,0 +1,48 @@
+# -*- coding: utf-8 -*-
+# pylint: disable=missing-docstring
+# pylint: disable=redefined-outer-name
+"""Tests for the entropy() function."""
+import numpy
+import pytest
+
+import ndd
+from inputs import Counts, Pmf
+from make_test_ref import approx
+
+
+@pytest.fixture
+def pmf():
+    return Pmf()
+
+
+@pytest.fixture
+def pmf_with_zeros():
+    return Pmf(zero=0.5)
+
+
+@pytest.fixture
+def counts():
+    return Counts()
+
+
+def test_pmf(pmf):
+    ref = pmf.entropy
+    assert ndd.entropy(pmf.pk) == approx(ref)
+
+
+def test_pmf_with_zeros(pmf_with_zeros):
+    ref = pmf_with_zeros.entropy
+    print(pmf_with_zeros.pk.sum())
+    assert ndd.entropy(pmf_with_zeros.pk) == approx(ref)
+
+
+def test_counts(counts):
+    estimator = ndd.estimators.AutoEstimator()
+    assert ndd.entropy(counts.nk) == approx(
+        counts.entropy(estimator=estimator))
+
+
+def test_unnormalized_pmf():
+    counts = numpy.random.random(size=100)  # pylint: disable=no-member
+    pk = counts / counts.sum()
+    assert ndd.entropy(counts) == approx(Pmf().entropy_from_pmf(pk))

tests/test_estimators.py

@@ -0,0 +1,21 @@
+# -*- coding: utf-8 -*-
+# pylint: disable=missing-docstring
+# pylint: disable=redefined-outer-name
+"""Estimators tests."""
+import pytest
+
+import ndd
+from inputs import Pmf
+from make_test_ref import approx
+
+
+@pytest.fixture
+def pmf():
+    return Pmf()
+
+
+def test_PmfPlugin(pmf):
+    """Test estimator from PMF."""
+    estimator = ndd.estimators.PmfPlugin()
+    ref = pmf.entropy
+    assert estimator(pmf.pk) == approx(ref)