GitList
Browse code

Merge pull request #274 from python-adaptive/feature/min_npoints

AverageLearner: implement min_npoints

Bas Nijholt authored on 14/05/2020 20:17:38 • GitHub committed on 14/05/2020 20:17:38
Showing 2 changed files
adaptive/learner/average_learner.py
History View file @ 1c60897
... ... 
@@ -19,6 +19,8 @@ class AverageLearner(BaseLearner):
19 19 
         Desired absolute tolerance.
20 20 
     rtol : float
21 21 
         Desired relative tolerance.
22 
+    min_npoints : int
23 
+        Minimum number of points to sample.
22 24
23 25 
     Attributes
24 26 
     ----------
... ... 
@@ -30,7 +32,7 @@ class AverageLearner(BaseLearner):
30 32 
         Number of evaluated points.
31 33 
     """
32 34
33 
-    def __init__(self, function, atol=None, rtol=None):
35 
+    def __init__(self, function, atol=None, rtol=None, min_npoints=2):
34 36 
         if atol is None and rtol is None:
35 37 
             raise Exception("At least one of `atol` and `rtol` should be set.")
36 38 
         if atol is None:
... ... 
@@ -44,6 +46,8 @@ class AverageLearner(BaseLearner):
44 46 
         self.atol = atol
45 47 
         self.rtol = rtol
46 48 
         self.npoints = 0
49 
+        # Cannot estimate standard deviation with fewer than 2 points.
50 
+        self.min_npoints = max(min_npoints, 2)
47 51 
         self.sum_f = 0
48 52 
         self.sum_f_sq = 0
49 53
... ... 
@@ -92,7 +96,7 @@ class AverageLearner(BaseLearner):
92 96 
         """The corrected sample standard deviation of the values
93 97 
         in `data`."""
94 98 
         n = self.npoints
95 
-        if n < 2:
99 
+        if n < self.min_npoints:
96 100 
             return np.inf
97 101 
         numerator = self.sum_f_sq - n * self.mean ** 2
98 102 
         if numerator < 0:
... ... 
@@ -106,7 +110,7 @@ class AverageLearner(BaseLearner):
106 110 
             n = self.npoints if real else self.n_requested
107 111 
         else:
108 112 
             n = n
109 
-        if n < 2:
113 
+        if n < self.min_npoints:
110 114 
             return np.inf
111 115 
         standard_error = self.std / sqrt(n)
112 116 
         return max(
... ... 
@@ -150,10 +154,11 @@ class AverageLearner(BaseLearner):
150 154 
             self.function,
151 155 
             self.atol,
152 156 
             self.rtol,
157 
+            self.min_npoints,
153 158 
             self._get_data(),
154 159 
         )
155 160
156 161 
     def __setstate__(self, state):
157 
-        function, atol, rtol, data = state
158 
-        self.__init__(function, atol, rtol)
162 
+        function, atol, rtol, min_npoints, data = state
163 
+        self.__init__(function, atol, rtol, min_npoints)
159 164 
         self._set_data(data)
adaptive/tests/test_average_learner.py
History View file @ 1c60897
... ... 
@@ -4,6 +4,7 @@ import flaky
4 4 
 import numpy as np
5 5
6 6 
 from adaptive.learner import AverageLearner
7 
+from adaptive.runner import simple
7 8
8 9
9 10 
 def test_only_returns_new_points():
... ... 
@@ -46,3 +47,15 @@ def test_avg_std_and_npoints():
46 47 
             assert learner.npoints == len(learner.data)
47 48 
             assert abs(learner.sum_f - values.sum()) < 1e-13
48 49 
             assert abs(learner.std - std) < 1e-13
50 
+
51 
+
52 
+def test_min_npoints():
53 
+    def constant_function(seed):
54 
+        return 0.1
55 
+
56 
+    for min_npoints in [1, 2, 3]:
57 
+        learner = AverageLearner(
58 
+            constant_function, atol=0.01, rtol=0.01, min_npoints=min_npoints
59 
+        )
60 
+        simple(learner, lambda l: l.loss() < 1)
61 
+        assert learner.npoints >= max(2, min_npoints)