search

ruizhang-ai / SIGN-Detecting-Beneficial-Feature-Interactions-for-Recommender-Systems

Detecting Beneficial Feature Interactions for Recommender Systems, AAAI 2021
MIT License
33 stars 6 forks source link

Experimental setup for comparison with baseline #4

Open Flyingkeyboard699 opened 2 years ago

Flyingkeyboard699 commented 2 years ago

This is an interesting work. When I realized the baseline of Frappe data set, I found that the evaluation metrics were better than those reported in this paper. I want to know whether your baseline hyperparameter setting is consistent with the model proposed in this paper or whether grid search is the best.

ethanmock commented 2 years ago

Hi,

We use the DeepCTR package (https://deepctr-doc.readthedocs.io/en/latest/index.html) to run the baseline models. The MLP structure and other settings are consistent with our model for a fair comparison.

Flyingkeyboard699 commented 2 years ago

Hi，ethanmock. Thanks for the reply. I found that the learning rate and regularization rate have a great impact on the final result. I want to know whether the two hyper-parameters settings of the baseline model refer to DeepCTR package or are consistent with those of your model.