Problem in Eval.py?

[AndRossi]

Hi, thanks for developing CapsE.
I think that using capsule networks may be the next big thing in Link prediction!

I'm opening this issue because I have the impression that your evaluation strategy, implemented in evalCapsE.py, may suffer from the same controversies emerged in ConvKB under this issue:
daiquocnguyen/ConvKB#5

Am I correct?

[dschaehi]

This problem is addressed in detail in the following paper:

Z. Sun, S. Vashishth, S. Sanyal, P. Talukdar, Y. Yang, A Re-evaluation of Knowledge Graph Completion Methods, ArXiv:1911.03903 [Cs]. (2019). http://arxiv.org/abs/1911.03903 (accessed November 14, 2019).

If the existence of the mentioned problem is true, then the authors should make this public somewhere (e.g., https://ml-retrospectives.github.io/). I think this is important to advance research in this area.

[daiquocnguyen]

I have answered this issue at daiquocnguyen/ConvKB#5