def test_uniform():
l = Symbol('l', real=True)
w = Symbol('w', positive=True, finite=True)
X = Uniform('x', l, l + w)
assert simplify(E(X)) == l + w/2
assert simplify(variance(X)) == w**2/12
# With numbers all is well
X = Uniform('x', 3, 5)
assert P(X < 3) == 0 and P(X > 5) == 0
assert P(X < 4) == P(X > 4) == Rational(1, 2)
def test_cdf():
X = Normal('x', 0, 1)
d = cdf(X)
assert P(X < 1) == d(1)
assert d(0) == Rational(1, 2)
d = cdf(X, X > 0) # given X>0
assert d(0) == 0
Y = Exponential('y', 10)
d = cdf(Y)
assert d(-5) == 0
assert P(Y > 3) == 1 - d(3)
pytest.raises(ValueError, lambda: cdf(X + Y))
Z = Exponential('z', 1)
f = cdf(Z)
z = Symbol('z')
assert f(z) == Piecewise((1 - exp(-z), z >= 0), (0, True))
U = Uniform('x', 3, 5)
u = cdf(U)
assert u(z) == z/2 - Rational(3, 2)
def test_uniform_P():
"""This stopped working because SingleContinuousPSpace.compute_density no
longer calls integrate on a DiracDelta but rather just solves directly.
integrate used to call UniformDistribution.expectation which special-cased
subsed out the Min and Max terms that Uniform produces.
I decided to regress on this class for general cleanliness (and I suspect
speed) of the algorithm.
"""
l = Symbol('l', real=True)
w = Symbol('w', positive=True, finite=True)
X = Uniform('x', l, l + w)
assert P(X < l) == 0 and P(X > l + w) == 0
def test_prefab_sampling():
N = Normal('X', 0, 1)
L = LogNormal('L', 0, 1)
E = Exponential('Ex', 1)
P = Pareto('P', 1, 3)
W = Weibull('W', 1, 1)
U = Uniform('U', 0, 1)
B = Beta('B', 2, 5)
G = Gamma('G', 1, 3)
variables = [N, L, E, P, W, U, B, G]
niter = 10
for var in variables:
for i in range(niter):
assert sample(var) in var.pspace.domain.set
