baichuan / disambiguation_embedding Goto Github PK

Thank you very much for sharing the code.
I read your paper and redid the experiment on your code. I have tried several times, but the result of several names in the Arnetminer dataset are much smaller than which in Table 3 of the paper.(e.g.,Bin Yu, Yang Wang, Yu Zhang and Wei Xu). I don't know if I missed something or had a problem with the parameter setting.
the parameter setting:
dimension: 20
learning rate: 0.02
regularization param: {0.001, 0.005, 0.01, 0.1, 1, 10}
epoch num:{ 100, 150, 200, 250, 300 }
I want to get close to the results of the paper. Hope for your advice or comments~