MPLP: Metapath-Based Label Propagation for Heterogenous Graphs

Related tags

Deep Learningmplp
Overview

MPLP: Metapath-Based Label Propagation for Heterogenous Graphs

Results on MAG240M

Here, we demonstrate the following performance on the MAG240M dataset from [email protected] 2021.

Model Test Acc Validation Acc Parameters Hardware
Our Model 0.7447 0.7669 ± 0.0003 (ensemble 0.7696) 743,449 Tesla V100 (21GB)

Reproducing results

0. Requirements

Here just list python3 packages we used in this competition:

numpy==1.19.2
torch==1.5.1+cu101
dgl-cu101==0.6.0.post1
ogb==1.3.1
sklearn==0.23.2
tqdm==4.46.1

1. Prepare Graph and Features

The preprocess code modifed from dgl baseline. We created graph with 6 different edge types instead of 5.

# Time cost: 3hours,30mins

python3 $MAG_CODE_PATH/preprocess.py
        --rootdir $MAG_INPUT_PATH \
        --author-output-path $MAG_PREP_PATH/author.npy \
        --inst-output-path $MAG_PREP_PATH/inst.npy \
        --graph-output-path $MAG_PREP_PATH \
        --graph-as-homogeneous \
        --full-output-path $MAG_PREP_PATH/full_feat.npy

The graphs and features will be saved in MAG_PREP_PATH , where the MAG_PREP_PATH is specified in run.sh.

Calculate features

The meta-path based features are generated by this script. Details can be found in our technical report.

# Time cost: 2hours,20mins (only generate label related features)

python3 $MAG_CODE_PATH/feature.py
        $MAG_INPUT_PATH \
        $MAG_PREP_PATH/dgl_graph_full_heterogeneous_csr.bin \
        $MAG_FEAT_PATH \
        --seed=42

Train RGAT model and prepare RGAT features

The RGAT model is modifed from dgl baseline. The validation accuracy is 0.701 , as same as described in the dgl baseline github.

# Time cost: 33hours,40mins (20mins for each epoch)

python3 $MAG_CODE_PATH/rgat.py
        --rootdir $MAG_INPUT_PATH \
        --graph-path $MAG_PREP_PATH/dgl_graph_full_homogeneous_csc.bin \
        --full-feature-path $MAG_PREP_PATH/full_feat.npy \
        --output-path $MAG_RGAT_PATH/ \
        --epochs=100 \
        --model-path $MAG_RGAT_PATH/model.pt \
        --submission-path $MAG_RGAT_PATH/

You will get embeddings as input features of the following MPLP models.

2. Train MPLP models

The train process splits to two steps:

  1. train the model with full train samples at a large learning rate (here we use StepLR(lr=0.01, step_size=100, gamma=0.25))
  2. then fine tune the model with latest train samples (eg, paper with year >= 2018) with a small learning rate (0.000625)

You can train the MPLP model by running the following commands:

# Time cost: 2hours,40mins for each seed

for seed in $(seq 0 7);
do
    python3 $MAG_CODE_PATH/mplp.py \
            $MAG_INPUT_PATH \
            $MAG_MPLP_PATH/data/ \
            $MAG_MPLP_PATH/output/seed${seed} \
            --gpu \
            --seed=${seed} \
            --batch_size=10240 \
            --epochs=200 \
            --num_layers=2 \
            --learning_rate=0.01 \
            --dropout=0.5 \
            --num_splits=5
done

3. Ensemble MPLP results

While having all the results with k-fold cross validation training under 8 different seeds, you can average the results by running code below:

python3 $MAG_CODE_PATH/ensemble.py $MAG_MPLP_PATH/output/ $MAG_SUBM_PATH
Owner
Qiuying Peng
Astrophysics -> Data Science
Qiuying Peng
GitHub Repository
用opencv的dnn模块做yolov5目标检测，包含C++和Python两个版本的程序

yolov5-dnn-cpp-py yolov5s,yolov5l,yolov5m,yolov5x的onnx文件在百度云盘下载， 链接：https://pan.baidu.com/s/1d67LUlOoPFQy0MV39gpJiw 提取码：bayj python版本的主程序是main_yolov5.

365 Jan 04, 2023
The official implementation of Theme Transformer

Theme Transformer This is the official implementation of Theme Transformer. Checkout our demo and paper : Demo | arXiv Environment: using python versi

Ian Shih 85 Dec 08, 2022
Bu repo SAHI uygulamasını mantığını öğreniyoruz.

SAHI-Learn: SAHI'den Beraber Kodlamak İster Misiniz Herkese merhabalar ben Kadir Nar. SAHI kütüphanesine gönüllü geliştiriciyim. Bu repo SAHI kütüphan

Kadir Nar 11 Aug 22, 2022
GAN encoders in PyTorch that could match PGGAN, StyleGAN v1/v2, and BigGAN. Code also integrates the implementation of these GANs.

MTV-TSA: Adaptable GAN Encoders for Image Reconstruction via Multi-type Latent Vectors with Two-scale Attentions. This is the official code release fo

owl 37 Dec 24, 2022
Official Pytorch Implementation of Adversarial Instance Augmentation for Building Change Detection in Remote Sensing Images.

IAug_CDNet Official Implementation of Adversarial Instance Augmentation for Building Change Detection in Remote Sensing Images. Overview We propose a

53 Dec 02, 2022
Locally Enhanced Self-Attention: Rethinking Self-Attention as Local and Context Terms

LESA Introduction This repository contains the official implementation of Locally Enhanced Self-Attention: Rethinking Self-Attention as Local and Cont

Chenglin Yang 20 Dec 31, 2021
Code Release for the paper "TriBERT: Full-body Human-centric Audio-visual Representation Learning for Visual Sound Separation"

TriBERT This repository contains the code for the NeurIPS 2021 paper titled "TriBERT: Full-body Human-centric Audio-visual Representation Learning for

UBC Computer Vision Group 8 Aug 31, 2022
EMNLP 2021 paper Models and Datasets for Cross-Lingual Summarisation.

This repository contains data and code for our EMNLP 2021 paper Models and Datasets for Cross-Lingual Summarisation. Please contact me at

9 Oct 28, 2022
On Nonlinear Latent Transformations for GAN-based Image Editing - PyTorch implementation

On Nonlinear Latent Transformations for GAN-based Image Editing - PyTorch implementation On Nonlinear Latent Transformations for GAN-based Image Editi

Valentin Khrulkov 22 Oct 24, 2022
Predicting Tweet Sentiment Maching Learning and streamlit

Predicting-Tweet-Sentiment-Maching-Learning-and-streamlit (I prefere using Visual Studio Code ) Open the folder in VS Code Run the first cell in requi

1 Nov 20, 2021
Google-drive-to-sqlite - Create a SQLite database containing metadata from Google Drive

google-drive-to-sqlite Create a SQLite database containing metadata from Google

Simon Willison 140 Dec 04, 2022
Dahua Camera and Doorbell Home Assistant Integration

Home Assistant Dahua Integration The Dahua Home Assistant integration allows you to integrate your Dahua cameras and doorbells in Home Assistant. It's

Ronnie 216 Dec 26, 2022
WeakVRD-Captioning - Implementation of paper Improving Image Captioning with Better Use of Caption

WeakVRD-Captioning - Implementation of paper Improving Image Captioning with Better Use of Caption

30 Oct 28, 2022
Official PyTorch implementation of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image", ICCV 2019

PoseNet of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image" Introduction This repo is official Py

Gyeongsik Moon 677 Dec 25, 2022
Deep Markov Factor Analysis (NeurIPS2021)

Deep Markov Factor Analysis (DMFA) Codes and experiments for deep Markov factor analysis (DMFA) model accepted for publication at NeurIPS2021: A. Farn

Sarah Ostadabbas 2 Dec 16, 2022
RGBD-Net - This repository contains a pytorch lightning implementation for the 3DV 2021 RGBD-Net paper.

[3DV 2021] We propose a new cascaded architecture for novel view synthesis, called RGBD-Net, which consists of two core components: a hierarchical depth regression network and a depth-aware generator

Phong Nguyen Ha 4 May 26, 2022
The code for our NeurIPS 2021 paper "Kernelized Heterogeneous Risk Minimization".

Kernelized-HRM Jiashuo Liu, Zheyuan Hu The code for our NeurIPS 2021 paper "Kernelized Heterogeneous Risk Minimization"[1]. This repo contains the cod

Liu Jiashuo 8 Nov 20, 2022
Fast Soft Color Segmentation

Fast Soft Color Segmentation

3 Oct 29, 2022
An essential implementation of BYOL in PyTorch + PyTorch Lightning

Essential BYOL A simple and complete implementation of Bootstrap your own latent: A new approach to self-supervised Learning in PyTorch + PyTorch Ligh

Enrico Fini 48 Sep 27, 2022
Zero-shot Synthesis with Group-Supervised Learning (ICLR 2021 paper)

GSL - Zero-shot Synthesis with Group-Supervised Learning Figure: Zero-shot synthesis performance of our method with different dataset (iLab-20M, RaFD,

Andy_Ge 62 Dec 21, 2022