add ensemble estimators

pydistinct/ensemble_estimators.py

@@ -0,0 +1,49 @@
+"""
+This is a module to be used as a reference for building other modules
+"""
+
+from pydistinct.stats_estimators import *
+
+
+def median_estimator(sequence):
+    """
+    Takes median result from all statistical estimators.
+
+    :param sequence: sample sequence of integers
+    :type sequence: array of ints
+    :return: median value of all estimator
+    :rtype: float
+    """
+    d = len(set(sequence))
+
+    def try_wrap(func, **kwargs):
+        try:
+            return func(**kwargs)
+        except:
+            return d
+
+    estimators = [("chao_estimator", try_wrap(chao_estimator, sequence=sequence)),
+                  ("chao_lee_estimator", try_wrap(chao_lee_estimator, sequence=sequence)),
+                  ("jackknife_estimator", try_wrap(jackknife_estimator, sequence=sequence)),
+                  ("sichel_estimator", try_wrap(sichel_estimator, sequence=sequence)),
+                  ("bootstrap_estimator", try_wrap(bootstrap_estimator, sequence=sequence)),
+                  ("method_of_moments_estimator", try_wrap(method_of_moments_estimator, sequence=sequence))
+                  ]
+
+    for n_pop in [1000, 10000000]:
+        estimators.append(("horvitz_thompson_estimator_{}".format(n_pop),
+                           try_wrap(horvitz_thompson_estimator, sequence=sequence,
+                                    pop_estimator=lambda x: x * n_pop)))
+        estimators.append(("method_of_moments_v2_estimator_{}".format(n_pop),
+                           try_wrap(method_of_moments_v2_estimator, sequence=sequence,
+                                    pop_estimator=lambda x: x * n_pop)))
+
+        estimators.append(("method_of_moments_v3_estimator_{}".format(n_pop),
+                           try_wrap(method_of_moments_v3_estimator, sequence=sequence,
+                                    pop_estimator=lambda x: x * n_pop)))
+
+        estimators.append(("smoothed_jackknife_estimator_{}".format(n_pop),
+                           try_wrap(smoothed_jackknife_estimator, sequence=sequence,
+                                    pop_estimator=lambda x: x * n_pop)))
+
+    return np.median(list(map(lambda x: x[1], estimators)))