LightGCN

Introduction

[paper]

Title: LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation

Authors: Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang, Meng Wang

Abstract: Graph Convolution Network (GCN) has become new state-of-the-art for collaborative filtering. Nevertheless, the reasons of its effectiveness for recommendation are not well understood. Existing work that adapts GCN to recommendation lacks thorough ablation analyses on GCN, which is originally designed for graph classification tasks and equipped with many neural network operations. However, we empirically find that the two most common designs in GCNs — feature transformation and nonlinear activation — contribute little to the performance of collaborative filtering. Even worse, including them adds to the difficulty of training and degrades recommendation performance.

In this work, we aim to simplify the design of GCN to make it more concise and appropriate for recommendation. We propose a new model named LightGCN, including only the most essential component in GCN — neighborhood aggregation — for collaborative filtering. Specifically, LightGCN learns user and item embeddings by linearly propagating them on the user-item interaction graph, and uses the weighted sum of the embeddings learned at all layers as the final embedding. Such simple, linear, and neat model is much easier to implement and train, exhibiting substantial improvements (about 16.0% relative improvement on average) over Neural Graph Collaborative Filtering (NGCF) — a state-of-the-art GCN-based recommender model — under exactly the same experimental setting. Further analyses are provided towards the rationality of the simple LightGCN from both analytical and empirical perspectives.

../../../_images/lightgcn.png

Running with RecBole

Model Hyper-Parameters:

  • embedding_size (int) : The embedding size of users and items. Defaults to 64.

  • n_layers (int) : The number of layers in LightGCN. Defaults to 2.

  • reg_weight (float) : The L2 regularization weight. Defaults to 1e-05.

A Running Example:

Write the following code to a python file, such as run.py

from recbole.quick_start import run_recbole

run_recbole(model='LightGCN', dataset='ml-100k')

And then:

python run.py

Tuning Hyper Parameters

If you want to use HyperTuning to tune hyper parameters of this model, you can copy the following settings and name it as hyper.test.

learning_rate choice [0.01,0.005,0.001,0.0005,0.0001]
n_layers choice [1,2,3,4]
reg_weight choice [1e-05,1e-04,1e-03,1e-02]

Note that we just provide these hyper parameter ranges for reference only, and we can not guarantee that they are the optimal range of this model.

Then, with the source code of RecBole (you can download it from GitHub), you can run the run_hyper.py to tuning:

python run_hyper.py --model=[model_name] --dataset=[dataset_name] --config_files=[config_files_path] --params_file=hyper.test

For more details about Parameter Tuning, refer to Parameter Tuning.

If you want to change parameters, dataset or evaluation settings, take a look at