def test_statistics(self):
self.a = Distribution(cyclic=True)
self.a.add({0.0: 0.0, pi / 2: 1.0})
self.assertAlmostEqual(vector_sum(self.a)[0], 1.0)
self.assertAlmostEqual(weighted_average(self.a), pi / 2)
self.assertAlmostEqual(vector_sum(self.a)[1], weighted_average(self.a))
self.assertAlmostEqual(selectivity(self.a), 1.0)
self.a.add({-pi / 2: 1.0}) # (should be like 3pi/2)
self.assertAlmostEqual(vector_sum(self.a)[0], 0.0)
self.assertAlmostEqual(weighted_average(self.a), 0.0)
self.assertAlmostEqual(selectivity(self.a), 0.0)
self.a.add({3 * pi / 8: 0.3})
self.assertAlmostEqual(vector_sum(self.a)[0], 0.3)
self.assertAlmostEqual(weighted_average(self.a), 3 * pi / 8)
self.assertAlmostEqual(selectivity(self.a),
vector_sum(self.a)[0] / 2.3)
self.c = Distribution((0.0, 1.0), cyclic=True)
self.c.add({0.0: 1.0, 0.25: 1.0})
self.assertAlmostEqual(vector_sum(self.c)[0], (1.0 + 1.0)**0.5)
self.assertAlmostEqual(weighted_average(self.c), vector_sum(self.c)[1])
self.assertEqual(vector_sum(self.c)[1], atan2(1.0, 1.0) / (2 * pi))
self.c.add({1.75: 1.0}) # added beyond bounds
self.assertAlmostEqual(vector_sum(self.c)[0], 1.0)
self.assertEqual(weighted_average(self.c), 0.0)
self.d = Distribution(axis_bounds=(0.0, 1.0),
cyclic=False,
keep_peak=False)
self.assertEqual(self.d.undefined_vals, 0)
self.assertAlmostEqual(selectivity(self.d), 1.0)
self.d.add({0.0: 0.0})
self.assertAlmostEqual(selectivity(self.d), 1.0)
self.d.add({0.5: 0.0})
self.assertAlmostEqual(selectivity(self.d), 0.0)
# note that selectivity() actually includes a call to DSF_WeightedAverage
self.assertEqual(self.d.undefined_vals, 4)
self.assertAlmostEqual(weighted_average(self.d), 0.0)
self.assertEqual(self.d.undefined_vals, 6)
self.d.add({0.0: 1.0})
self.assertAlmostEqual(selectivity(self.d), 1.0)
self.assertEqual(self.d.undefined_vals, 6)
self.d.add({0.5: 1.0})
self.assertAlmostEqual(selectivity(self.d), 0.0)
self.d.add({0.75: 2.0})
self.assertAlmostEqual(selectivity(self.d), 0.25)
def setUp(self):
self.h = Distribution((0.0, 5.0))
self.h.add({0: 0.0})
self.h.add({1: 0.1})
self.h.add({2: 0.2})
self.h.add({2: 0.2})
self.h.add({3: 0.3})
self.h.add({3: 0.3})
self.h.add({3: 0.3})
self.h.add({4: 0.4})
self.h.add({4: 0.4})
self.h.add({4: 0.4})
self.h.add({4: 0.4})
self.h.add({5: 0.5})
self.h.add({5: 0.5})
self.h.add({5: 0.5})
self.h.add({5: 0.5})
self.h.add({5: 0.5})
self.g = copy.deepcopy(self.h)
self.g.add({0: 0.0})
self.g.add({1: 0.1})
self.g.add({2: 0.2})
self.g.add({3: 0.3})
self.g.add({4: 0.4})
self.g.add({5: 0.5})
def test_lissom_statistics(self):
self.assertEqual(len(self.h.values()), 6)
self.assertAlmostEqual(self.h.total_value, 5.5)
self.assertEqual(self.h.total_count, 16)
self.assertAlmostEqual(max(self.h.values()), 2.5)
self.assertAlmostEqual(min(self.h.values()), 0.0)
self.assertEqual(max(self.h.counts()), 5)
self.assertEqual(min(self.h.counts()), 1)
self.assertAlmostEqual(self.g.total_value, 7.0)
self.assertEqual(self.g.total_count, 22)
self.assertAlmostEqual(sum(self.g.values()), 7.0)
self.assertAlmostEqual(self.g.weighted_sum(), 28.0)
# (should return _weighted_average:)
self.assertAlmostEqual(weighted_average(self.g),
28.0 / sum(self.g.values()))
self.assertAlmostEqual(vector_sum(self.g)[1], 4.40449904283)
self.q = Distribution((0, 4), cyclic=True)
self.q.add({3: 1})
self.q.add({0: 0, 1: 0, 2: 0, 4: 0})
self.assertAlmostEqual(vector_sum(self.q)[0], 1.0)
self.assertAlmostEqual(weighted_average(self.q), 3.0)
self.assertAlmostEqual(vector_sum(self.q)[1], 3.0)
# Example where this matches LISSOM by using an empty bin at 5.
self.rr = Distribution(
(0, 5), cyclic=True
) # 5 because in the L. test example 0 and 4 are distinct
self.rr.add({0: 1, 1: 1, 2: 1, 3: 1, 4: 1})
self.assertAlmostEqual(vector_sum(self.rr)[0], 0.0)
self.assertAlmostEqual(weighted_average(self.rr),
vector_sum(self.rr)[1])
self.rr.add({1: 2})
self.assertAlmostEqual(vector_sum(self.rr)[0], 2.0)
self.assertAlmostEqual(weighted_average(self.rr), 1.0)
self.rr.add({3: 2})
self.assertAlmostEqual(vector_sum(self.rr)[0], 2 * 2 * cos(2 * pi / 5))
self.assertAlmostEqual(weighted_average(self.rr), 2.0)
def test_lissom_peak_histogram(self):
self.p = Distribution((0, 2), keep_peak=True)
self.p.add({0: 1.0})
self.p.add({2: 9.9})
self.p.add({2: 3.3})
self.p.add({0: 1.0})
self.p.add({1: 0.3})
self.p.add({1: 84.0})
self.p.add({1: 36.0})
self.assertAlmostEqual(self.p.get_value(0), 1.0)
self.assertAlmostEqual(self.p.get_value(1), 84.0)
self.assertAlmostEqual(self.p.get_value(2), 9.9)
self.assertEqual(self.p.get_count(0), 2)
self.assertEqual(self.p.get_count(1), 3)
self.assertEqual(self.p.get_count(2), 2)
