Apoorv Saxena

Can you please elaborate on what experiment you did, what commands you used?

Thanks for the details, let me try it with this config and I'll get back to you

Hi, I will try too add ASAP

@namadjidku Relation matching was needed for only 3-hop full setting in MetaQA since the paths are too long for the base model to handle properly. For 3-hop half, relation matching...

Not yet unfortunately. I will pin this issue until it gets added. However, based on more recent work in this area/MetaQA dataset, I would recommend against using EmbedKGQA in full...

We did, and all multiplicative models performed similarly for us. We chose ComplEx because it was slightly better.

You might want to look into this paper: https://openreview.net/pdf?id=BkxSmlBFvr People have pointed out that rather than the scoring function, the training scheme (eg. negative sampling, loss) matters more.

@ShuangNYU What command did you use, and what was the exact output?

The point they make in the paper is that all scoring functions perform roughly the same. The reason that earlier models reported lesser numbers than modern reproduction is mostly due...

Oh got it. Did you do this in the full KG setting? If so, I think the performance would be same for all models since to answer questions you need...