search

claws-lab / jodie

A PyTorch implementation of ACM SIGKDD 2019 paper "Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks"
MIT License
355 stars 82 forks source link

without features same performance #11

Open snash4 opened 4 years ago

snash4 commented 4 years ago

Hi, In one of our experiments we excluded features, we get exactly the same results for Reddit and slightly better (MRR=0.759 and Recall@10=0.824) for Wikipedia. So does that mean that features don't have impact on the learning?

Thanks.