Skip to content

Code for CIKM2020 "S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization"

Notifications You must be signed in to change notification settings

aHuiWang/CIKM2020-S3Rec

Repository files navigation

Code for our CIKM 2020 Paper "S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization"

Overview

avatar

Reproduce

Please check the ./reproduce/ directory.

Results

Performance comparison of different methods on six datasets. The best performance and the second best performance methods are denoted in bold and underlined fonts respectively.

In the PAPER, we pair the ground-truth item with 99 randomly sampled negative items that the user has not interacted with, and report the results of HR@{1, 5, 10}, NDCG@{5, 10} and MRR. The used test files are name as

data-name_sample.txt

The results are shown in the following pic. avatar

We alse rank the ground-truth item with all the items. We omit the FM and AutoInt because they need enumerate all user-item pairs, which take a very long time.

The results are shown in the following pic.

avatar

requirements

pip install -r requirements.txt

data format

data preprocess
./data/data_process.py

generate negative items for testing
./data/generate_test.py


data-name.txt
one user per line
user_1 item_1 item_2 ...
user_2 item_1 item_2 ...

data-name_sample.txt
one user per line
user_1 neg_item_1 neg_item_2 ...
user_2 neg_item_1 neg_item_2 ...

data-name_item2attributes.json
{item_1:[attr, ...], item_2:[attr, ...], ... }

pretrain

python run_pretrain.py \
--data_name data_name

finetune

We support two evaluation methods. For more details, please check the ./reproduce directory.

  • Rank ground-truth item with 99 randomly sampled negative items
python run_finetune_sample.py \
--data_name data_name \
--ckp pretrain_epochs_num
  • Rank the ground-truth item with all the items
python run_finetune_full.py \
--data_name data_name \
--ckp pretrain_epochs_num

Cite

If you find the our codes and datasets useful for your research or development, please cite our paper:

@inproceedings{DBLP:conf/cikm/ZhouWZZWZWW20,
  author    = {Kun Zhou and
               Hui Wang and
               Wayne Xin Zhao and
               Yutao Zhu and
               Sirui Wang and
               Fuzheng Zhang and
               Zhongyuan Wang and
               Ji{-}Rong Wen},
  title     = {S3-Rec: Self-Supervised Learning for Sequential Recommendation with
               Mutual Information Maximization},
  booktitle = {{CIKM} '20: The 29th {ACM} International Conference on Information
               and Knowledge Management, Virtual Event, Ireland, October 19-23, 2020},
  pages     = {1893--1902},
  publisher = {{ACM}},
  year      = {2020}
}

Contact

If you have any question for our paper or codes, please send email to hui.wang@ruc.edu.cn.

About

Code for CIKM2020 "S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization"

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages