def test_sample(self, tpe):
"""Test GMMSampler sample function"""
mus = numpy.linspace(-3, 3, num=12, endpoint=False)
sigmas = [0.5] * 12
gmm_sampler = GMMSampler(tpe, mus, sigmas, -3, 3)
points = gmm_sampler.sample(25)
points = numpy.array(points)
assert len(points) <= 25
assert numpy.all(points >= -3)
assert numpy.all(points < 3)
mus = numpy.linspace(-10, 10, num=10, endpoint=False)
sigmas = [0.00001] * 10
weights = numpy.linspace(1, 10, num=10)**3
numpy.random.shuffle(weights)
weights = weights / weights.sum()
gmm_sampler = GMMSampler(tpe, mus, sigmas, -11, 9, weights)
points = gmm_sampler.sample(10000)
points = numpy.array(points)
hist = numpy.histogram(points,
bins=[-11, -9, -7, -5, -3, -1, 1, 3, 5, 7, 9])
assert numpy.all(hist[0].argsort() == numpy.array(weights).argsort())
assert numpy.all(points >= -11)
assert numpy.all(points < 9)
def test_sample_narrow_space(self, tpe):
"""Test that sampling in a narrow space does not fail to fast"""
mus = numpy.ones(12) * 0.5
sigmas = [0.5] * 12
times = []
for bounds in [(0.4, 0.6), (0.49, 0.51), (0.499, 0.501), (0.49999, 0.50001)]:
gmm_sampler = GMMSampler(tpe, mus, sigmas, *bounds)
times.append(timeit.timeit(lambda: gmm_sampler.sample(2), number=100))
# Test that easy sampling takes less time.
assert sorted(times) == times
gmm_sampler = GMMSampler(
tpe, mus, sigmas, 0.05, 0.04, attempts_factor=1, max_attempts=10
)
with pytest.raises(RuntimeError) as exc:
gmm_sampler.sample(1, attempts=10)
assert exc.match("Failed to sample in interval")
def test_get_loglikelis(self):
"""Test to get log likelis of points"""
mus = numpy.linspace(-10, 10, num=10, endpoint=False)
weights = numpy.linspace(1, 10, num=10)**3
numpy.random.shuffle(weights)
weights = weights / weights.sum()
sigmas = [0.00001] * 10
gmm_sampler = GMMSampler(tpe, mus, sigmas, -11, 9, weights)
points = [mus[7]]
pdf = norm(mus[7], sigmas[7])
point_likeli = numpy.log(pdf.pdf(mus[7]) * weights[7])
likelis = gmm_sampler.get_loglikelis(points)
assert list(likelis) == point_likeli
assert likelis[0] == point_likeli
sigmas = [2] * 10
gmm_sampler = GMMSampler(tpe, mus, sigmas, -11, 9, weights)
log_pdf = []
pdfs = []
for i in range(10):
pdfs.append(norm(mus[i], sigmas[i]))
for pdf, weight in zip(pdfs, weights):
log_pdf.append(numpy.log(pdf.pdf(0) * weight))
point_likeli = numpy.log(numpy.sum(numpy.exp(log_pdf)))
points = numpy.random.uniform(-11, 9, 30)
points = numpy.insert(points, 10, 0)
likelis = gmm_sampler.get_loglikelis(points)
point_likeli = numpy.format_float_scientific(point_likeli,
precision=10)
gmm_likeli = numpy.format_float_scientific(likelis[10], precision=10)
assert point_likeli == gmm_likeli
assert len(likelis) == len(points)