Could scikit-learn label propagation be used with non-negative samples?

Thanks for your introduction, as I was looking for a method to classify positive and unlabeled data for long.
I have tried svmlin and got 80.1% accuracy.
Now I want to find another method for result comparison, but when I read the document of scikit-learn label propagation, I found it wouldn't work if there isn't any negative sample at all. Is it true?
