[SIGIR 2023] This is the official PyTorch implementation for the paper: "EulerNet: Adaptive Feature Interaction Learning via Euler’s Formula for CTR Prediction".
The architecture of our proposed EulerNet
tensorflow==2.4.1
python==3.7.3
cudatoolkit==11.3.1
pytorch==1.11.0
We follow FmFM to process the Criteo and Avazu Dataset, and follow DCNV2 to process the MovieLens-1M dataset. The scripts for dataset processing can be found under the /DataSource folder. You first need to download the raw dataset files and put them into the /DataSource folder.
Then pre-process the data:
python DataSource/[dataset]_parse.py
Finally, get the files for training, validation, and testing:
python DataSource/split.py
python train.py --config_files=[dataset].yaml
If you find EulerNet useful for your research or development, please cite the following papers: EulerNet.
@inproceedings{tian2023eulernet,
title = {EulerNet: Adaptive Feature Interaction Learning via Euler's Formula for CTR Prediction},
author = {Tian, Zhen and Bai, Ting and Zhao, Wayne Xin and Wen, Ji-Rong and Cao, Zhao},
booktitle = {Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages = {1376–1385},
year = {2023},
}