MNT Use cimport numpy as cnp for sklearn/decomposition (scikit-learn#…

…23330)

sklearn/decomposition/_online_lda_fast.pyx

@@ -1,22 +1,22 @@
 cimport cython
-cimport numpy as np
+cimport numpy as cnp
 import numpy as np
 
-np.import_array()
+cnp.import_array()
 
 from libc.math cimport exp, fabs, log
 from numpy.math cimport EULER
 
 
-def mean_change(np.ndarray[ndim=1, dtype=np.float64_t] arr_1,
- np.ndarray[ndim=1, dtype=np.float64_t] arr_2):
+def mean_change(cnp.ndarray[ndim=1, dtype=cnp.float64_t] arr_1,
+ cnp.ndarray[ndim=1, dtype=cnp.float64_t] arr_2):
  """Calculate the mean difference between two arrays.
 
  Equivalent to np.abs(arr_1 - arr2).mean().
  """
 
- cdef np.float64_t total, diff
- cdef np.npy_intp i, size
+ cdef cnp.float64_t total, diff
+ cdef cnp.npy_intp i, size
 
  size = arr_1.shape[0]
  total = 0.0
@@ -27,9 +27,9 @@ def mean_change(np.ndarray[ndim=1, dtype=np.float64_t] arr_1,
  return total / size
 
 
-def _dirichlet_expectation_1d(np.ndarray[ndim=1, dtype=np.float64_t] doc_topic,
+def _dirichlet_expectation_1d(cnp.ndarray[ndim=1, dtype=cnp.float64_t] doc_topic,
  double doc_topic_prior,
- np.ndarray[ndim=1, dtype=np.float64_t] out):
+ cnp.ndarray[ndim=1, dtype=cnp.float64_t] out):
  """Dirichlet expectation for a single sample:
  exp(E[log(theta)]) for theta ~ Dir(doc_topic)
  after adding doc_topic_prior to doc_topic, in-place.
@@ -39,8 +39,8 @@ def _dirichlet_expectation_1d(np.ndarray[ndim=1, dtype=np.float64_t] doc_topic,
  out[:] = np.exp(psi(doc_topic) - psi(np.sum(doc_topic)))
  """
 
- cdef np.float64_t dt, psi_total, total
- cdef np.npy_intp i, size
+ cdef cnp.float64_t dt, psi_total, total
+ cdef cnp.npy_intp i, size
 
  size = doc_topic.shape[0]
 
@@ -55,7 +55,7 @@ def _dirichlet_expectation_1d(np.ndarray[ndim=1, dtype=np.float64_t] doc_topic,
  out[i] = exp(psi(doc_topic[i]) - psi_total)
 
 
-def _dirichlet_expectation_2d(np.ndarray[ndim=2, dtype=np.float64_t] arr):
+def _dirichlet_expectation_2d(cnp.ndarray[ndim=2, dtype=cnp.float64_t] arr):
  """Dirichlet expectation for multiple samples:
  E[log(theta)] for theta ~ Dir(arr).
 
@@ -64,9 +64,9 @@ def _dirichlet_expectation_2d(np.ndarray[ndim=2, dtype=np.float64_t] arr):
  Note that unlike _dirichlet_expectation_1d, this function doesn't compute
  the exp and doesn't add in the prior.
  """
- cdef np.float64_t row_total, psi_row_total
- cdef np.ndarray[ndim=2, dtype=np.float64_t] d_exp
- cdef np.npy_intp i, j, n_rows, n_cols
+ cdef cnp.float64_t row_total, psi_row_total
+ cdef cnp.ndarray[ndim=2, dtype=cnp.float64_t] d_exp
+ cdef cnp.npy_intp i, j, n_rows, n_cols
 
  n_rows = arr.shape[0]
  n_cols = arr.shape[1]