#################################################

我们在原作者的开源代码基础上复现了原论文。 同时在另一数据集movielens-20M上分别使用GMF和NeuMF模型，使用不降噪、R_CE降噪和T_CE降噪的方式进行训练， 其结果与该论文的结论一致，使用R_CE和T_CE降噪后能有效提升模型精度。 此外，我们添加了CE，用于进行不做降噪处理的推荐器训练，与R_CE或T_CE的运行方式一致。

We reproduced the original paper based on the authors' open-source code. Additionally, we applied the GMF and NeuMF models on another dataset, MovieLens-20M, and trained them using no denoising, R_CE denoising, and T_CE denoising methods. The results were consistent with the conclusions of the original paper, showing that R_CE and T_CE denoising effectively improved the model accuracy. Furthermore, we added a CE method for training the recommender system without any denoising.

###################################################

！！值得注意的是！！ 在使用我们提供的数据集movielens-20M作实验时，应先合并以下文件： " data\movielens\movielens_train\movielens.train.rating.part* " 为 " data\movielens\movielens.train.rating"

参考方式： copy /b data\movielens\movielens_train\movielens.train.rating.part* data\movielens\merged.movielens.train.rating 或者使用python脚本等其他方式

When conducting experiments using our provided dataset, movielens-20M, you should first merge the following files: " data\movielens\movielens_train\movielens.train.rating.part* " into " data\movielens\movielens.train.rating "

Reference method: " copy /b data\movielens\movielens_train\movielens.train.rating.part* data\movielens\merged.movielens.train.rating " or use a Python script or other methods. #################################################

Adaptive Denoising Training for Recommendation.

This is the pytorch implementation of our paper at WSDM 2021:

Denoising Implicit Feedback for Recommendation.
Wenjie Wang, Fuli Feng, Xiangnan He, Liqiang Nie, Tat-Seng Chua.

• Anaconda 3
• python 3.7.3
• pytorch 1.4.0
• numpy 1.16.4

For others, please refer to the file env.yaml.

python main.py --dataset=$1 --model=$2 --drop_rate=$3 --num_gradual=$4 --gpu=$5

or use run.sh

sh run.sh dataset model drop_rate num_gradual gpu_id

The output will be in the ./log/xxx folder.

sh run.sh dataset model alpha gpu_id

We provide the code to inference based on the well-trained model parameters.

python inference.py --dataset=$1 --model=$2 --drop_rate=$3 --num_gradual=$4 --gpu=$5

1. Train GMF by T_CE on Yelp:

python main.py --dataset=yelp --model=GMF --drop_rate=0.1 --num_gradual=30000 --gpu=0

2. Train NeuMF by R_CE on Amazon_book

python main.py --dataset=amazon_book --model=NeuMF-end --alpha=_0.25 --gpu=0

We release all training logs in ./log folder. The hyperparameter settings can be found in the log file. The well-trained parameter files are too big to upload to Github. I will upload to drives later and share it here.

If you use our code, please kindly cite:

@inproceedings{wang2021denoising,
  title={Denoising implicit feedback for recommendation},
  author={Wang, Wenjie and Feng, Fuli and He, Xiangnan and Nie, Liqiang and Chua, Tat-Seng},
  booktitle={Proceedings of the 14th ACM international conference on web search and data mining},
  pages={373--381},
  publisher={ACM},
  year={2021}
}

Thanks to the NCF implementation:

• Tensorflow Version from Xiangnan He.
• Torch Version from Yangyang Guo.

Besides, this research is supported by the National Research Foundation, Singapore under its International Research Centres in Singapore Funding Initiative, and the National Natural Science Foundation of China (61972372, U19A2079). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of National Research Foundation, Singapore.

NUS © NExT++