Pre-Training Graph Neural Networks for Cold-Start Users and Items Representation.

Last update: Nov 14, 2022

Related tags

Overview

Pretrain-Recsys

This is our Tensorflow implementation for our WSDM 2021 paper:

Bowen Hao, Jing Zhang, Hongzhi Yin, Cuiping Li, Hong Chen. Pre-Training Graph Neural Networks for Cold-Start Users and Items Representation.

Environment Requirement The code has been tested running under Python 3.6.12. The required packages are as follows:

tensorflow == 1.14.0 numpy == 1.16.0 tqdm == 4.50.2 scipy == 1.5.2 torch == 1.3.1

Runing instructions: We test our model on three datasets, namely MovieLens-1M, MOOCs and Last.fm, and each dataset contains a folder. Each folder contrains five GNNs models, in which three models are contained in the GeneralConv.py, and two other models are contained in the FastGCN folder and the FBNE folder.

Before running our model, please download the oracle embedding files from https://pan.baidu.com/s/1XAR9thNt8PTHSvPQ8P2p-g (the password is hwj3), and put the oracle embedding files in the corresponding folders. For example, you can download the mooc oracle embedding file, and put it in the mooc folder.

For the pre-traning process, run ' python GeneralConv.py ', which consists of three GNNs pre-training models (GraphSAGE, GAT and LightGCN) or ' python FastGCNConv.py ' in the FastGCN folder or ' python FBNEConv.py ' in the FBNE folder. For the downstream recommendation process, run ' python downstream_trainer.py ' in the basic folder or the FastGCN/FBNE folder.

Pre-Training Graph Neural Networks for Cold-Start Users and Items Representation.

Related tags

Overview

Pretrain-Recsys

Owner

Pytorch Implementation of Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmetic

The Unsupervised Reinforcement Learning Benchmark (URLB)

A Closer Look at Invalid Action Masking in Policy Gradient Algorithms

Towers of Babel: Combining Images, Language, and 3D Geometry for Learning Multimodal Vision. ICCV 2021.

Code for the paper "Implicit Representations of Meaning in Neural Language Models"

Allows including an action inside another action (by preprocessing the Yaml file). This is how composite actions should have worked.

Face Depixelizer based on "PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models" repository.

We envision models that are pre-trained on a vast range of domain-relevant tasks to become key for molecule property prediction

Deep deconfounded recommender (Deep-Deconf) for paper "Deep causal reasoning for recommendations"

Python tools for 3D face: 3DMM, Mesh processing(transform, camera, light, render), 3D face representations.

Graph Self-Attention Network for Learning Spatial-Temporal Interaction Representation in Autonomous Driving

Picasso: A CUDA-based Library for Deep Learning over 3D Meshes

Code for our paper A Transformer-Based Feature Segmentation and Region Alignment Method For UAV-View Geo-Localization,

An LSTM based GAN for Human motion synthesis

Paper list of log-based anomaly detection

Codes to calculate solar-sensor zenith and azimuth angles directly from hyperspectral images collected by UAV. Works only for UAVs that have high resolution GNSS/IMU unit.

Automatically erase objects in the video, such as logo, text, etc.

Pytorch implementation of

(NeurIPS 2021) Realistic Evaluation of Transductive Few-Shot Learning

basic tutorial on pytorch