Graph embeddings via matrix factorization for link prediction: smoothing or truncating negatives?

Open in new window