Invalid Value Encountered in Sqrt

Nobody has claimed this yet.

I'm trying to apply UMAP to a relatively small dataset (412 x 80) but I keep getting the following error:

```
/Users/nlubock/miniconda2/envs/py36/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt
return distances if squared else np.sqrt(distances, out=distances)
```

Interestingly enough, this same error also appears in the Read the Docs in the `n_components` section.

I'm currently using Python 3.6.3 with:

```
numpy 1.14.5
pandas 0.23.3
scikit-learn 0.19.1
scipy 1.1.0
numba 0.38.1
umap-learn 0.3.0
```

Here's [the data](https://github.com/lmcinnes/umap/files/2211068/full-mat.csv.zip) in case it's something to do with that.

Reproduce the warning with the linked full-mat.csv data and the reported Python, NumPy, scikit-learn, SciPy, numba, and umap-learn versions. Start at scikit-learn's pairwise.py sqrt warning and the Read the Docs n_components section, then determine whether the dataset or version interaction causes it. Done means identifying a reproducible cause and documenting or correcting the affected behavior.

Written by the indexing model from the issue text.