Official implementation of NeurIPS'21: Implicit SVD for Graph Representation Learning

Last update: Jan 08, 2023

Related tags

Overview

isvd

Official implementation of NeurIPS'21: Implicit SVD for Graph Representation Learning

If you find this code useful, you may cite us as:

@inproceedings{haija2021isvd,
  author={Sami Abu-El-Haija AND Hesham Mostafa AND Marcel Nassar AND Valentino Crespi AND Greg Ver Steeg AND Aram Galstyan},
  title={Implicit SVD for Graph Representation Learning},
  booktitle={Advances in Neural Information Processing Systems},
  year={2021},
}

To run link prediction on Stanford SNAP and node2vec datasets:

To embed with rank-32 SVD:

python3 run_snap_linkpred.py --dataset_name=ppi --dim=32
python3 run_snap_linkpred.py --dataset_name=ca-AstroPh --dim=32
python3 run_snap_linkpred.py --dataset_name=ca-HepTh --dim=32
python3 run_snap_linkpred.py --dataset_name=soc-facebook --dim=32

To embed with rank 256 on half of the training edges, determine "best rank" based on the remaining half, then re-run sVD with the best rank on all of training: (note: negative dim causes this logic):

python3 run_snap_linkpred.py --dataset_name=ppi --dim=-256
python3 run_snap_linkpred.py --dataset_name=ca-AstroPh --dim=-256
python3 run_snap_linkpred.py --dataset_name=ca-HepTh --dim=-256
python3 run_snap_linkpred.py --dataset_name=soc-facebook --dim=-256

To run semi-supervised node classification on Planetoid datasets

You must first download the planetoid dataset as:

mkdir -p ~/data
cd ~/data
git clone [email protected]:kimiyoung/planetoid.git

Afterwards, you may navigate back to this directory and run our code as:

python3 run_planetoid.py --dataset=ind.citeseer
python3 run_planetoid.py --dataset=ind.cora
python3 run_planetoid.py --dataset=ind.pubmed

To run link prediction on Stanford OGB DDI

python3 ogb_linkpred_sing_val_net.py

Note the above will download the dataset from Stanford. If you already have it, you may symlink it into directory dataset

To run link prediction on Stanford OGB ArXiv

As our code imports gttf, you must first clone it onto the repo:

git clone [email protected]:isi-usc-edu/gttf.git

Afterwards, you may run as:

python3 final_obgn_mixed_device.py --funetune_device='gpu:0'

Note the above will download the dataset from Stanford. If you already have it, you may symlink it into directory dataset. You may skip the finetune_device argument if you do not have a GPU installed.

Official implementation of NeurIPS'21: Implicit SVD for Graph Representation Learning

Related tags

Overview

isvd

To run link prediction on Stanford SNAP and node2vec datasets:

To run semi-supervised node classification on Planetoid datasets

To run link prediction on Stanford OGB DDI

To run link prediction on Stanford OGB ArXiv

Owner

Sami Abu-El-Haija

Few-Shot Graph Learning for Molecular Property Prediction

In-Place Activated BatchNorm for Memory-Optimized Training of DNNs

moving object detection for satellite videos.

Multi-Joint dynamics with Contact. A general purpose physics simulator.

A Pytorch implementation of "LegoNet: Efficient Convolutional Neural Networks with Lego Filters" (ICML 2019).

RM Operation can equivalently convert ResNet to VGG, which is better for pruning; and can help RepVGG perform better when the depth is large.

The source code and data of the paper "Instance-wise Graph-based Framework for Multivariate Time Series Forecasting".

Multi-Scale Aligned Distillation for Low-Resolution Detection (CVPR2021)

Differential rendering based motion capture blender project.

CondLaneNet: a Top-to-down Lane Detection Framework Based on Conditional Convolution

Vision-Language Pre-training for Image Captioning and Question Answering

MMdnn is a set of tools to help users inter-operate among different deep learning frameworks. E.g. model conversion and visualization. Convert models between Caffe, Keras, MXNet, Tensorflow, CNTK, PyTorch Onnx and CoreML.

PyTorchMemTracer - Depict GPU memory footprint during DNN training of PyTorch

Implementation of the GVP-Transformer, which was used in the paper "Learning inverse folding from millions of predicted structures" for de novo protein design alongside Alphafold2

Performance Analysis of Multi-user NOMA Wireless-Powered mMTC Networks: A Stochastic Geometry Approach

Pytorch implementation of "Get To The Point: Summarization with Pointer-Generator Networks"

Official Tensorflow implementation of U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation (ICLR 2020)

Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021)

Few-shot Learning of GPT-3

Mitsuba 2: A Retargetable Forward and Inverse Renderer