OverRec

Code for ICDM 2022 paper, Beyond Double Ascent via Recurrent Neural Tangent Kernel in Sequential Recommendation.

Usage

Download datasets from RecSysDatasets or their Google Drive. And put the files in ./dataset/ like the following.

$ tree
.
├── Amazon_Beauty
│   ├── Amazon_Beauty.inter
│   └── Amazon_Beauty.item
├── Amazon_Clothing_Shoes_and_Jewelry
│   ├── Amazon_Clothing_Shoes_and_Jewelry.inter
│   └── Amazon_Clothing_Shoes_and_Jewelry.item
├── Amazon_Sports_and_Outdoors
│   ├── Amazon_Sports_and_Outdoors.inter
│   └── Amazon_Sports_and_Outdoors.item
├── ml-1m
│   ├── ml-1m.inter
│   ├── ml-1m.item
│   ├── ml-1m.user
│   └── README.md
└── yelp
    ├── README.md
    ├── yelp.inter
    ├── yelp.item
    └── yelp.user

Run python3 run_overrec.py.

This experiment does not require GPU calculation. It is purely CPU computation. Yet it may require large memory for kernel calculation, ranging from 100 Gb ~ 300 Gb for different datasets.

MISC

We implement SKNN and STAN in this repo. Run python3 run_sknn.py or python3 run_stan.py.

Cite

If you find this repo useful, please cite

@article{OverRec,
  author    = {Ruihong Qiu and
               Zi Huang and
               Hongzhi Yin},
  title     = {Beyond Double Ascent via Recurrent Neural Tangent Kernel in Sequential Recommendation},
  journal   = {CoRR},
  volume    = {abs/2209.03735},
  year      = {2022},
}

Credit

This repo is based on RecBole and RNTK