the method of embedding table initialization for different CTR prediction models

reczoo / BARS

BARS: Towards Open Benchmarking for Recommender Systems https://openbenchmark.github.io/BARS

Apache License 2.0

342 stars 55 forks source link

the method of embedding table initialization for different CTR prediction models #27

Closed Zeng-B-B closed 7 months ago

Zeng-B-B commented 9 months ago

How to choose the method of embedding table Initialization? if I choose the xavier_uniform method, the AUC is always 0.5 when I run the code of SAM model.But when I run the code of other models (AutoInt, FiGNN, AFN)using the xavier_uniform method, the results are good and normal. I see that the method using in the project is normal initialization(embedding_initializer="partial(nn.init.normal_, std=1e-4)"). Is all model the mehod?

zhujiem commented 7 months ago

Yes, basically we did not tune the embedding_initializer. It is normal by default.