qianrenjian / EulerNet

This is the official PyTorch implementation for the paper: "EulerNet: Adaptive Feature Interaction Learning via Euler’s Formula for CTR Prediction".

EulerNet

[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

Requirements

tensorflow==2.4.1
python==3.7.3
cudatoolkit==11.3.1
pytorch==1.11.0

Dataset Preparation

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

Training

python train.py --config_files=[dataset].yaml