[MRG] Add error message when X passed to silhouette score is 1d (fixes #6190)

[yenchenlin]

I've added an error message to tell user that first argument passed into silhouette_score cannot be 1d.

Solve #6190
Can @amueller please review this?

[yenchenlin]

I've tested this implementation with the following script:

import timeit
import numpy as np
from sklearn.metrics.cluster.unsupervised import silhouette_score

def test_silhouette_score():
    random_state = np.random.RandomState(seed=0)
    random_clusters = random_state.randint(low=0, high=2, size=(10 ** 3, 100))
    random_clusters2 = random_state.randint(low=0, high=2, size=10 ** 3)

    silhouette_score(random_clusters, random_clusters2)

time = timeit.Timer(test_silhouette_score)
print min(time.repeat(number=10 ** 3))

This script calls silhouette_score 1000 times with input data which consists of 1000 samples and 100 features.

original version prints 51.3936669827 s,
while the modified version prints 52.0506219864 s,
which means it barely affects the performance.

[MechCoder]

You should do X = np.asarray(X) here and check to avoid another array conversion inside check_X_y if X is a list etc

[MechCoder]

LGTM except 2 nitpicks

[yenchenlin]

Hello @MechCoder ,
Thanks for your opinion.

I've modified the code, can you please have a look?

[amueller]

Do it after check_X_y and you don't have to call asarray. I was hoping there was an option to check_X_y to enforce this, but it doesn't look like this.

[yenchenlin]

Close since it is fixed.