Duplicated triples generated in train_test_split_no_unseen()

Accenture / AmpliGraph

Python library for Representation Learning on Knowledge Graphs https://docs.ampligraph.org

Apache License 2.0

2.14k stars 251 forks source link

Closed pyvandenbussche closed 5 years ago

pyvandenbussche commented 5 years ago

The script to carve a train/test set from a list of triples leads to:

duplicated test triples
and therefore the train set being the diff of all triples - test ones is bigger than expected.

Will do a pull request with fix.

chanlevan commented 5 years ago

Reopen to make sure the results must be deterministic and there should be a unit test to cover.