def test_variance_condition():
base_conditions = [
MaxEntropyCondition(weight=0.001),
SmoothnessCondition(),
IntervalCondition(p=0.95, min=0.3, max=0.7),
]
base_dist = PointDensity.from_conditions(base_conditions,
verbose=True,
scale=Scale(0, 1))
base_variance = base_dist.variance()
increased_variance = base_variance + 0.01
# Increase in variance should decrease peak
var_condition = VarianceCondition(variance=increased_variance, weight=1)
var_conditions = base_conditions + [var_condition]
var_dist = PointDensity.from_conditions(var_conditions,
verbose=True,
scale=Scale(0, 1))
assert np.max(var_dist.normed_densities) < np.max(
base_dist.normed_densities)
# Highly weighted variance condition should make var very close to specified var
strong_condition = VarianceCondition(variance=increased_variance,
weight=100000)
strong_var_conditions = base_conditions + [strong_condition]
strong_var_dist = PointDensity.from_conditions(strong_var_conditions,
verbose=True,
scale=Scale(0, 1))
assert strong_var_dist.variance() == pytest.approx(
float(increased_variance), abs=0.001)
def test_variance_condition():
def get_variance(dist):
xs = np.linspace(dist.scale.low, dist.scale.high, dist.ps.size)
mean = np.dot(dist.ps, xs)
return np.dot(dist.ps, np.square(xs - mean))
base_conditions = [
MaxEntropyCondition(weight=0.1),
SmoothnessCondition(),
IntervalCondition(p=0.95, min=0.3, max=0.7),
]
base_dist = HistogramDist.from_conditions(base_conditions, verbose=True)
base_variance = get_variance(base_dist)
increased_variance = base_variance + 0.01
# Increase in variance should decrease peak
var_condition = VarianceCondition(variance=increased_variance, weight=1)
var_conditions = base_conditions + [var_condition]
var_dist = HistogramDist.from_conditions(var_conditions, verbose=True)
assert np.max(var_dist.ps) < np.max(base_dist.ps)
# Highly weighted variance condition should make var very close to specified var
strong_condition = VarianceCondition(variance=increased_variance,
weight=1000)
strong_var_conditions = base_conditions + [strong_condition]
strong_var_dist = HistogramDist.from_conditions(strong_var_conditions,
verbose=True)
assert get_variance(strong_var_dist) == pytest.approx(
float(increased_variance), abs=0.001)
def test_interval_plus_entropy(scale: Scale):
conditions = [
IntervalCondition(p=0.5, max=scale.denormalize_point(0.3)),
MaxEntropyCondition(weight=0.01),
]
fitted_dist = PointDensity.from_conditions(
conditions,
scale=scale,
)
# We expect at most 3 different densities: one for inside the interval, one for outside,
# and one between.
assert np.unique(fitted_dist.normed_densities).size <= 3
def test_density_cdf(scale: Scale):
uniform_dist = PointDensity.from_conditions([MaxEntropyCondition()],
scale=scale)
# Off of scale
assert uniform_dist.cdf(scale.denormalize_point(-0.5)) == 0
assert uniform_dist.cdf(scale.denormalize_point(1.5)) == 1
# Edges of scale
assert uniform_dist.cdf(scale.denormalize_point(0.005)) < uniform_dist.cdf(
scale.denormalize_point(0.015))
assert uniform_dist.cdf(scale.denormalize_point(0.985)) < uniform_dist.cdf(
scale.denormalize_point(0.995))
# Denormalized onto a different scale
denormalized_dist = uniform_dist.denormalize(Scale(0, 2))
assert denormalized_dist.cdf(1) == pytest.approx(0.5, abs=0.01)
assert denormalized_dist.cdf(1.5) != 0
assert denormalized_dist.cdf(2.5) == 1
def test_mean_condition():
def get_mean(dist):
xs = np.linspace(dist.scale.low, dist.scale.high, dist.ps.size)
return np.dot(dist.ps, xs)
base_conditions = [MaxEntropyCondition(weight=0.1)]
base_dist = HistogramDist.from_conditions(base_conditions, verbose=True)
base_mean = get_mean(base_dist)
# Mean condition should move mean closer to specified mean
mean_conditions = base_conditions + [MeanCondition(mean=0.25, weight=1)]
mean_dist = HistogramDist.from_conditions(mean_conditions, verbose=True)
assert abs(get_mean(mean_dist) - 0.25) < abs(base_mean - 0.25)
# Highly weighted mean condition should make mean very close to specified mean
strong_condition = MeanCondition(mean=0.25, weight=1000)
strong_mean_conditions = base_conditions + [strong_condition]
strong_mean_dist = HistogramDist.from_conditions(strong_mean_conditions,
verbose=True)
assert get_mean(strong_mean_dist) == pytest.approx(0.25, rel=0.01)
def test_mean_condition():
base_conditions = [MaxEntropyCondition(weight=0.1)]
base_dist = PointDensity.from_conditions(base_conditions,
verbose=True,
scale=Scale(0, 1))
base_mean = base_dist.mean()
# Mean condition should move mean closer to specified mean
mean_conditions = base_conditions + [MeanCondition(mean=0.25, weight=1)]
mean_dist = PointDensity.from_conditions(mean_conditions,
verbose=True,
scale=Scale(0, 1))
assert abs(mean_dist.mean() - 0.25) < abs(base_mean - 0.25)
# Highly weighted mean condition should make mean very close to specified mean
strong_condition = MeanCondition(mean=0.25, weight=100000)
strong_mean_conditions = base_conditions + [strong_condition]
strong_mean_dist = PointDensity.from_conditions(strong_mean_conditions,
verbose=True,
scale=Scale(0, 1))
assert strong_mean_dist.mean() == pytest.approx(0.25, rel=0.01)
def test_fit_hist_regression_1():
"""
Regression test for bug: "This custom question has a weird histogram - why?"
see https://elicit.ought.org/builder/gflpsSBAb
for more on the bug, see
https://docs.google.com/document/d/1CFklTKtbKzXi6-lRaEsX4ZiY3Yzpbfdg7i2j1NvKP34/edit#heading=h.ph1huakxn33f
"""
conditions = [
IntervalCondition(p=0.25, max=2.0),
IntervalCondition(p=0.75, max=4.0),
IntervalCondition(p=0.9, max=6.0),
MaxEntropyCondition(weight=0.1),
]
histogram_dist = HistogramDist.from_conditions(conditions,
scale=Scale(low=0, high=52))
assert histogram_dist.cdf(2) == pytest.approx(0.25, abs=0.05)
assert histogram_dist.ppf(0.9) == pytest.approx(6, abs=1)
def test_density_pdf(scale: Scale):
uniform_dist = PointDensity.from_conditions([MaxEntropyCondition()],
scale=scale)
assert uniform_dist.pdf(scale.denormalize_point(0.5)) != 0
assert uniform_dist.pdf(scale.denormalize_point(1.5)) == 0
def test_hist_pdf(scale: Scale):
uniform_dist = HistogramDist.from_conditions([MaxEntropyCondition()], scale=scale)
assert uniform_dist.pdf(scale.denormalize_point(0.5)) != 0
assert uniform_dist.pdf(scale.denormalize_point(1.5)) == 0