sklearn.metrics.silhouette_samples does not work with sparse matrices #18524
Description
Describe the bug
silhouette_samples(X, y, metric="precomputed") fails with ValueError: diag requires an array of at least two dimensions if X is sparse.
Steps/Code to Reproduce
from sklearn.metrics import silhouette_samples
from sklearn.neighbors import kneighbors_graph
from sklearn.datasets import make_blobs
X, y = make_blobs(n_samples=1000, centers=5, n_features=10)
pdist = kneighbors_graph(X, 5, mode='distance')
# pdist is scipy.sparse.csr.csr_matrix
silhouette_samples(pdist, y, metric="precomputed")Expected Results
No error is thrown.
Actual Results
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-14-140890024e6c> in <module>
----> 1 silhouette_samples(pdist, y, metric="precomputed")
/data1/mschroeder/miniconda3/envs/20-assdc/lib/python3.7/site-packages/sklearn/metrics/cluster/_unsupervised.py in silhouette_samples(X, labels, metric, **kwds)
216 if metric == 'precomputed':
217 atol = np.finfo(X.dtype).eps * 100
--> 218 if np.any(np.abs(np.diagonal(X)) > atol):
219 raise ValueError(
220 'The precomputed distance matrix contains non-zero '
<__array_function__ internals> in diagonal(*args, **kwargs)
/data1/mschroeder/miniconda3/envs/20-assdc/lib/python3.7/site-packages/numpy/core/fromnumeric.py in diagonal(a, offset, axis1, axis2)
1615 return asarray(a).diagonal(offset=offset, axis1=axis1, axis2=axis2)
1616 else:
-> 1617 return asanyarray(a).diagonal(offset=offset, axis1=axis1, axis2=axis2)
1618
1619
ValueError: diag requires an array of at least two dimensions
Possible fix
Change np.diagonal(X) to X.diagonal(), because this is implemented for numpy.ndarray as well as scipy.sparse.csr_matrix.diagonal. Or should this rather be fixed in numpy (so that np.diagonal does the right thing for sparse matrices)?
Versions
System:
python: 3.7.6 | packaged by conda-forge | (default, Jun 1 2020, 18:57:50) [GCC 7.5.0]
executable: /data1/mschroeder/miniconda3/envs/20-assdc/bin/python
machine: Linux-4.4.0-109-generic-x86_64-with-debian-stretch-sid
Python dependencies:
pip: 20.1.1
setuptools: 47.3.1.post20200616
sklearn: 0.22.2.post1
numpy: 1.18.5
scipy: 1.4.1
Cython: None
pandas: 1.0.5
matplotlib: 3.2.1
joblib: 0.15.1
Built with OpenMP: True