Infinite Recommendation Networks (∞-AE)

This repository contains the implementation of ∞-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, ∞-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

The paper also proposes Distill-CF: how to use ∞-AE for data distillation to create terse, high-fidelity, and synthetic data summaries for model training. We provide Distill-CF's code in a separate GitHub repository.

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

@article{inf_ae_distill_cf,
  title={Infinite Recommendation Networks: A Data-Centric Approach},
  author={Sachdeva, Noveen and Dhaliwal, Mehak Preet and Wu, Carole-Jean and McAuley, Julian},
  booktitle={Advances in Neural Information Processing Systems},
  series={NeurIPS '22},
  year={2022}
}

Code Author: Noveen Sachdeva ([email protected])

---

Setup

Environment Setup

pip install -r requirements.txt

Data Setup

This repository already includes the pre-processed data for ML-1M, Amazon Magazine, and Douban datasets as described in the paper. The code for pre-processing is in preprocess.py.

---

How to train ∞-AE?

• Edit the hyper_params.py file which lists all config parameters of ∞-AE.
• Finally, type the following command to train and evaluate ∞-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	∞-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: ∞-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