def test_batch_exec_2(self):
from scipy.stats import multivariate_normal
def pdf(X, Y):
std = 1
ndim_y = Y.shape[1]
return multivariate_normal.pdf(Y,
mean=np.zeros(ndim_y),
cov=np.eye(ndim_y) * std**2)
n_queries = 8 * 10**4
X = np.ones((n_queries, 2)) * 2
Y = np.stack([
np.linspace(-3, 3, num=n_queries),
np.linspace(-3, 3, num=n_queries)
],
axis=-1)
p_true = pdf(X, Y)
p_batched = execute_batch_async_pdf(pdf,
X,
Y,
batch_size=10000,
n_jobs=8)
self.assertLessEqual(np.mean((p_true - p_batched)**2), 0.00001)
def cdf(self, X, Y):
""" Predicts the conditional cumulative probability p(Y<=y|X=x). Requires the model to be fitted.
Args:
X: numpy array to be conditioned on - shape: (n_samples, n_dim_x)
Y: numpy array of y targets - shape: (n_samples, n_dim_y)
Returns:
conditional cumulative probability p(Y<=y|X=x) - numpy array of shape (n_query_samples, )
"""
assert self.fitted, "model must be fitted to compute likelihood score"
X, Y = self._handle_input_dimensionality(X, Y)
n_samples = X.shape[0]
if n_samples > MULTIPROC_THRESHOLD:
execute_batch_async_pdf(self.sm_kde.cdf, Y, X, n_jobs=self.n_jobs)
else:
return self.sm_kde.cdf(endog_predict=Y, exog_predict=X)
def test_batch_exec_1(self):
def pdf(X, Y):
return Y[:, 0]
n_queries = 10**3
X = np.ones((n_queries, 2)) * 2
Y = np.stack([
np.linspace(-3, 3, num=n_queries),
np.linspace(-3, 3, num=n_queries)
],
axis=-1)
p_true = pdf(X, Y)
p_batched = execute_batch_async_pdf(pdf, X, Y, batch_size=10000)
self.assertLessEqual(np.mean((p_true - p_batched)**2), 0.00001)
def log_pdf(self, X, Y):
""" Predicts the conditional log-probability log p(y|x). Requires the model to be fitted.
Args:
X: numpy array to be conditioned on - shape: (n_samples, n_dim_x)
Y: numpy array of y targets - shape: (n_samples, n_dim_y)
Returns:
conditional log-probability log p(y|x) - numpy array of shape (n_query_samples, )
"""
X, Y = self._handle_input_dimensionality(X, Y, fitting=True)
n_samples = X.shape[0]
if n_samples >= _MULTIPROC_THRESHOLD:
return execute_batch_async_pdf(self._log_pdf, X, Y, n_jobs=self.n_jobs)
else:
return self._log_pdf(X, Y)
def pdf(self, X, Y):
""" Predicts the conditional density p(y|x). Requires the model to be fitted.
Args:
X: numpy array to be conditioned on - shape: (n_samples, n_dim_x)
Y: numpy array of y targets - shape: (n_samples, n_dim_y)
Returns:
conditional probability density p(y|x) - numpy array of shape (n_query_samples, )
"""
assert self.fitted, "model must be fitted for predictions"
X, Y = self._handle_input_dimensionality(X, Y)
n_samples = X.shape[0]
if n_samples >= MULTIPROC_THRESHOLD:
return execute_batch_async_pdf(self._pdf, X, Y, n_jobs=self.n_jobs)
else:
return self._pdf(X, Y)
def pdf(self, X, Y):
""" Predicts the conditional likelihood p(y|x). Requires the model to be fitted.
Args:
X: numpy array to be conditioned on - shape: (n_samples, n_dim_x)
Y: numpy array of y targets - shape: (n_samples, n_dim_y)
Returns:
conditional likelihood p(y|x) - numpy array of shape (n_query_samples, )
"""
X, Y = self._handle_input_dimensionality(X, Y)
n_samples = X.shape[0]
if n_samples >= MULTIPROC_THRESHOLD:
return execute_batch_async_pdf(self.sm_kde.pdf,
Y,
X,
n_jobs=self.n_jobs)
else:
return self.sm_kde.pdf(endog_predict=Y, exog_predict=X)