79619600

Date: 2025-05-13 12:06:04
Score: 2.5
Natty:
Report link

Managed to find an answer to my own question : negative sampling was poorly done (mainly random links) which led the model to always have the same percentage of trust for positive and negative links. I coded my own neg sampling function and made it generate links only inside the said "gamme". Now i have around 0.88 AUC.

Reasons:
  • No code block (0.5):
  • Self-answer (0.5):
  • Single line (0.5):
  • Low reputation (1):
Posted by: lili