Skip to content

MAINT Remove -Wcpp warnings when compiling sklearn.manifold._utils #24925

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Nov 30, 2022
Merged

MAINT Remove -Wcpp warnings when compiling sklearn.manifold._utils #24925

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
8b33492
remove wcpp warnings
rprkh Nov 15, 2022
064fbff
include changes
rprkh Nov 16, 2022
2a43a8f
remove astype from return P
rprkh Nov 16, 2022
6505d3d
return base attribute of memoryview
rprkh Nov 21, 2022
08d3661
include TODO comment
rprkh Nov 22, 2022
3314cc9
keep previous behavior
jeremiedbb Nov 30, 2022
32206a6
Merge remote-tracking branch 'upstream/main' into pr/rprkh/24925
jeremiedbb Nov 30, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,6 +90,7 @@
"sklearn.linear_model._sag_fast",
"sklearn.linear_model._sgd_fast",
"sklearn.manifold._barnes_hut_tsne",
"sklearn.manifold._utils",
"sklearn.metrics.cluster._expected_mutual_info_fast",
"sklearn.metrics._pairwise_distances_reduction._datasets_pair",
"sklearn.metrics._pairwise_distances_reduction._middle_term_computer",
Expand Down
14 changes: 7 additions & 7 deletions sklearn/manifold/_utils.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,6 @@ from libc cimport math
import numpy as np
cimport numpy as cnp

cnp.import_array()


cdef extern from "numpy/npy_math.h":
Expand All @@ -12,8 +11,9 @@ cdef extern from "numpy/npy_math.h":
cdef float EPSILON_DBL = 1e-8
cdef float PERPLEXITY_TOLERANCE = 1e-5

cpdef cnp.ndarray[cnp.float32_t, ndim=2] _binary_search_perplexity(
cnp.ndarray[cnp.float32_t, ndim=2] sqdistances,
# TODO: have this function support float32 and float64 and preserve inputs' dtypes.
def _binary_search_perplexity(
const cnp.float32_t[:, :] sqdistances,
float desired_perplexity,
int verbose):
"""Binary search for sigmas of conditional Gaussians.
Expand All @@ -23,7 +23,7 @@ cpdef cnp.ndarray[cnp.float32_t, ndim=2] _binary_search_perplexity(

Parameters
----------
sqdistances : array-like, shape (n_samples, n_neighbors)
sqdistances : ndarray of shape (n_samples, n_neighbors), dtype=np.float32
Distances between training samples and their k nearest neighbors.
When using the exact method, this is a square (n_samples, n_samples)
distance matrix. The TSNE default metric is "euclidean" which is
Expand All @@ -37,7 +37,7 @@ cpdef cnp.ndarray[cnp.float32_t, ndim=2] _binary_search_perplexity(

Returns
-------
P : array, shape (n_samples, n_samples)
P : ndarray of shape (n_samples, n_samples), dtype=np.float64
Probabilities of conditional Gaussian distributions p_i|j.
"""
# Maximum number of binary search steps
Expand All @@ -63,7 +63,7 @@ cpdef cnp.ndarray[cnp.float32_t, ndim=2] _binary_search_perplexity(

# This array is later used as a 32bit array. It has multiple intermediate
# floating point additions that benefit from the extra precision
cdef cnp.ndarray[cnp.float64_t, ndim=2] P = np.zeros(
cdef cnp.float64_t[:, :] P = np.zeros(
(n_samples, n_neighbors), dtype=np.float64)

for i in range(n_samples):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You need to change the return line by:

return P.astype(np.float32)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not the current behavior. Currently it returns a float64 array (this is because it's using the def part of the function which doesn't care about the return type)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd rather keep the current behavior

Expand Down Expand Up @@ -118,4 +118,4 @@ cpdef cnp.ndarray[cnp.float32_t, ndim=2] _binary_search_perplexity(
if verbose:
print("[t-SNE] Mean sigma: %f"
% np.mean(math.sqrt(n_samples / beta_sum)))
return P
return np.asarray(P)