def test_distributions_mgd_random_sample():
mu = numpy.array([5., 1.])
cov = numpy.eye(2)
d = MultivariateGaussianDistribution(mu, cov)
x = numpy.array([[5.441227, 0.66913], [7.430771, 0.747908],
[5.10961, 2.582481]])
assert_array_almost_equal(d.sample(3, random_state=5), x)
assert_raises(AssertionError, assert_array_almost_equal, d.sample(5), x)
def test_distributions_mgd_random_sample(): mu = numpy.array([5., 1.]) cov = numpy.eye(2) d = MultivariateGaussianDistribution(mu, cov) x = numpy.array([[5.441227, 0.66913 ], [7.430771, 0.747908], [5.10961 , 2.582481]]) assert_array_almost_equal(d.sample(3, random_state=5), x) assert_raises(AssertionError, assert_array_almost_equal, d.sample(5), x)
def sample(self, n=10):
try:
X1 = np.stack(self.model.sample(int(np.round(0.8 * n))))
# np.round 取最近的整数
# 从多元高斯分布中采样8个点
if self.dim == 2:
mean = np.mean(X1, axis=0)
# 由X1的采样点计算均值
std = np.diag([1.0, 1.0])
gmm = MultivariateGaussianDistribution(mean, std)
# 构建一个新的多元高斯分布
X2 = np.stack(gmm.sample(int(np.round(0.1 * n))))
# 采样1个点
mean = np.mean(X1, axis=0)
std = np.diag([1e-3, 1e-3]) # 标准差不同了
gmm = MultivariateGaussianDistribution(mean, std)
X3 = np.stack(gmm.sample(int(np.round(0.1 * n))))
# 又采样了一个点
else:
mean = np.mean(X1, axis=0)
std = np.diag([1.0, 1.0, 1e-3])
gmm = MultivariateGaussianDistribution(mean, std)
X2 = np.stack(gmm.sample(int(np.round(0.1 * n))))
mean = np.mean(X1, axis=0)
std = np.diag([1e-3, 1e-3, 10.0])
gmm = MultivariateGaussianDistribution(mean, std)
X3 = np.stack(gmm.sample(int(np.round(0.1 * n))))
X = np.concatenate((X1, X2, X3)) # 多尺度采样？
except ValueError:
print("exception caught on sampling")
if self.dim == 2:
mean = np.zeros(self.dim)
std = np.diag([1.0, 1.0])
gmm = MultivariateGaussianDistribution(mean, std)
X = gmm.sample(int(n))
else:
mean = np.zeros(self.dim)
std = np.diag([1.0, 1.0, 5.0])
gmm = MultivariateGaussianDistribution(mean, std)
X = gmm.sample(int(n))
return X
def sample(self, n=10):
try:
X1 = np.stack(self.model.sample(int(np.round(0.8 * n))))
if self.dim == 2:
mean = np.mean(X1, axis=0)
std = np.diag([1.0, 1.0])
gmm = MultivariateGaussianDistribution(mean, std)
X2 = np.stack(gmm.sample(int(np.round(0.1 * n))))
mean = np.mean(X1, axis=0)
std = np.diag([1e-3, 1e-3])
gmm = MultivariateGaussianDistribution(mean, std)
X3 = np.stack(gmm.sample(int(np.round(0.1 * n))))
else:
mean = np.mean(X1, axis=0)
std = np.diag([1.0, 1.0, 1e-3])
gmm = MultivariateGaussianDistribution(mean, std)
X2 = np.stack(gmm.sample(int(np.round(0.1 * n))))
mean = np.mean(X1, axis=0)
std = np.diag([1e-3, 1e-3, 10.0])
gmm = MultivariateGaussianDistribution(mean, std)
X3 = np.stack(gmm.sample(int(np.round(0.1 * n))))
X = np.concatenate((X1, X2, X3))
except ValueError:
print("exception caught on sampling")
if self.dim == 2:
mean = np.zeros(self.dim)
std = np.diag([1.0, 1.0])
gmm = MultivariateGaussianDistribution(mean, std)
X = gmm.sample(int(n))
else:
mean = np.zeros(self.dim)
std = np.diag([1.0, 1.0, 5.0])
gmm = MultivariateGaussianDistribution(mean, std)
X = gmm.sample(int(n))
return X