Infinite Recommendation Networks ($\infty$-AE)

This repository contains the implementation of $\infty$-AE from the paper "Infinite Recommendation Networks: A Data-Centric Approach" [arXiv] where we leverage the NTK of an infinitely-wide autoencoder for implicit-feedback recommendation. Notably, $\infty$-AE:

• Is easy to implement (<50 lines of relevant code)
• Has a closed-form solution
• Has only a single hyper-parameter, $\lambda$
• Even though simplistic, outperforms all complicated SoTA models

If you find any module of this repository helpful for your own research, please consider citing the below under-review paper. Thanks!

@article{sachdeva2022b,
  title={Infinite Recommendation Networks: A Data-Centric Approach},
  author={Sachdeva, Noveen and Dhaliwal, Mehak Preet and Wu, Carole-Jean and McAuley, Julian},
  journal={arXiv preprint arXiv:2206.XXXXX},
  year={2022}
}

Code Author: Noveen Sachdeva ([email protected])

---

Setup

Environment Setup

pip install -r requirements.txt

Data Setup

Once you've correctly setup the python environment, the following script will download the ML-1M dataset and preprocess it for usage:

./setup.sh

---

How to train $\infty$-AE?

• Edit the hyper_params.py file which lists all config parameters of $\infty$-AE. Note that $\lambda$ is currently grid-searched in main.py so changing it will bring no difference, until the code is adjusted.
• Finally, type the following command to train and evaluate $\infty$-AE:

CUDA_VISIBLE_DEVICES=0 python main.py

---

Results sneak-peak

Below are the nDCG@10 results for the datasets used in the paper:

Dataset	PopRec	MF	NeuMF	MVAE	LightGCN	EASE	$\infty$-AE
Amazon Magazine	8.42	13.1	13.6	12.18	22.57	22.84	23.06
MovieLens-1M	13.84	25.65	24.44	22.14	28.85	29.88	32.82
Douban	11.63	13.21	13.33	16.17	16.68	19.48	24.94
Netflix	12.34	12.04	11.48	20.85	Timed out	26.83	30.59*

Note: $\infty$-AE's results on the Netflix dataset (marked with a *) are obtained by training only on 5% of the total users. Note however, all other methods are trained on the full dataset.

---

MIT License