def test_choice(weights, rng):
n = 1000
choices = [[1, 1], [1, -1], [-1, 1], [-1, -1]]
N = len(choices)
dist = Choice(choices, weights=weights)
# If d is passed, it has to match
with pytest.raises(ValueError):
dist.sample(n, d=4, rng=rng)
sample = dist.sample(n, rng=rng)
tsample, tchoices = list(map(tuple, sample)), list(map(tuple, choices))
# check that frequency of choices matches weights
inds = [tchoices.index(s) for s in tsample]
hist, bins = np.histogram(inds, bins=np.linspace(-0.5, N - 0.5, N + 1))
p_empirical = hist / float(hist.sum())
p = np.ones(N) / N if dist.p is None else dist.p
sterr = 1. / np.sqrt(n) # expected maximum standard error
assert np.allclose(p, p_empirical, atol=2 * sterr)
def test_choice(weights, rng, allclose):
"""Tests the choice function with weights"""
n = 1000
choices = [[1, 1], [1, -1], [-1, 1], [-1, -1]]
N = len(choices)
dist = Choice(choices, weights=weights)
# If d is passed, it has to match
with pytest.raises(ValueError):
dist.sample(n, d=4, rng=rng)
sample = dist.sample(n, rng=rng)
tsample = [tuple(point) for point in sample]
tchoices = [tuple(choice) for choice in choices]
# check that frequency of choices matches weights
inds = [tchoices.index(s) for s in tsample]
histogram, _ = np.histogram(inds, bins=np.linspace(-0.5, N - 0.5, N + 1))
p_empirical = histogram / float(histogram.sum())
p = np.ones(N) / N if dist.p is None else dist.p
sterr = 1.0 / np.sqrt(n) # expected maximum standard error
assert allclose(p, p_empirical, atol=2 * sterr)