Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods

Overview

Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods

Introduction

Graph Neural Networks (GNNs) have demonstrated superior performance in node classification or regression tasks, and have emerged as the state of the art in several applications. However, (inductive) GNNs require the edge connectivity structure of nodes to be known beforehand to work well. This is often not the case in several practical applications where the node degrees have power-law distributions, and nodes with a few connections might have noisy edges. An extreme case is the strict cold start (SCS) problem, where there is no neighborhood information available, forcing prediction models to rely completely on node features only. To study the viability of using inductive GNNs to solve the SCS problem, we introduce feature-contribution ratio (FCR), a metric to quantify the contribution of a node's features and that of its neighborhood in predicting node labels, and use this new metric as a model selection reward. We then propose Cold Brew, a new method that generalizes GNNs better in the SCS setting compared to pointwise and graph-based models, via a distillation approach. We show experimentally how FCR allows us to disentangle the contributions of various components of graph datasets, and demonstrate the superior performance of Cold Brew on several public benchmarks

Motivation

Long tail distribution is ubiquitously existed in large scale graph mining tasks. In some applications, some cold start nodes have too few or no neighborhood in the graph, which make graph based methods sub-optimal due to insufficient high quality edges to perform message passing.

gnns

gnns

Method

We improve teacher GNN with Structural Embedding, and propose student MLP model with latent neighborhood discovery step. We also propose a metric called FCR to judge the difficulty in cold start generalization.

gnns

coldbrew

Installation Guide

The following commands are used for installing key dependencies; other can be directly installed via pip or conda. A full redundant dependency list is in requirements.txt

pip install dgl
pip3 install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html
pip install torch-scatter -f https://pytorch-geometric.com/whl/torch-1.9.0+cu111.html
pip install torch-sparse -f https://pytorch-geometric.com/whl/torch-1.9.0+cu111.html
pip install torch-geometric

Training Guide

In options/base_options.py, a full list of useable args is present, with default arguments and candidates initialized.

Comparing between traditional GCN (optimized with Initial/Jumping/Dense/PairNorm/NodeNorm/GroupNorm/Dropouts) and Cold Brew's GNN (optimized with Structural Embedding)

Train optimized traditional GNN:

python main.py --dataset='Cora' --train_which='TeacherGNN' --whetherHasSE='000' --want_headtail=1 --num_layers=2 --use_special_split=1 Result: 84.15

python main.py --dataset='Citeseer' --train_which='TeacherGNN' --whetherHasSE='000' --want_headtail=1 --num_layers=2 --use_special_split=1 Result: 71.00

python main.py --dataset='Pubmed' --train_which='TeacherGNN' --whetherHasSE='000' --want_headtail=1 --num_layers=2 --use_special_split=1 Result: 78.2

Training Cold Brew's Teacher GNN:

python main.py --dataset='Cora' --train_which='TeacherGNN' --whetherHasSE='100' --se_reg=32 --want_headtail=1 --num_layers=2 --use_special_split=1 Result: 85.10

python main.py --dataset='Citeseer' --train_which='TeacherGNN' --whetherHasSE='100' --se_reg=0.5 --want_headtail=1 --num_layers=2 --use_special_split=1 Result: 71.40

python main.py --dataset='Pubmed' --train_which='TeacherGNN' --whetherHasSE='111' --se_reg=0.5 --want_headtail=1 --num_layers=2 --use_special_split=1 Result: 78.2

Comparing between MLP models:

Training naive MLP:

python main.py --dataset='Cora' --train_which='StudentBaseMLP' Result on isolation split: 63.92

Training GraphMLP:

python main.py --dataset='Cora' --train_which='GraphMLP' Result on isolation split: 68.63

Training Cold Brew's MLP:

python main.py --dataset='Cora' --train_which="SEMLP" --SEMLP_topK_2_replace=3 --SEMLP_part1_arch="2layer" --dropout_MLP=0.5 --studentMLP__opt_lr='torch.optim.Adam&0.005' Result on isolation split: 69.57

Hyperparameter meanings

--whetherHasSE: whether cold brew's TeacherGNN has structural embedding. The first ‘1’ means structural embedding exist in first layer; second ‘1’ means structural embedding exist in every middle layers; third ‘1’ means last layer.

--se_reg: regularization coefficient for cold brew teacher model's structural embedding.

--SEMLP_topK_2_replace: the number of top K best virtual neighbor nodes.

--manual_assign_GPU: set the GPU ID to train on. default=-9999, which means to dynamically choose GPU with most remaining memory.

Adaptation Guide

How to leverage this repo to train on other datasets:

In trainer.py, put any new graph dataset (node classification) under load_data() and return it.

what to load: return a dataset, which is a namespace, called 'data', data.x: 2D tensor, on cpu; shape = [N_nodes, dim_feature]. data.y: 1D tensor, on cpu; shape = [N_nodes]; values are integers, indicating the class of nodes. data.edge_index: tensor: [2, N_edge], cpu; edges contain self loop. data.train_mask: bool tensor, shape = [N_nodes], indicating the training node set. Template class for the 'data':

class MyDataset(torch_geometric.data.data.Data):
    def __init__(self):
        super().__init__()

Citation

comming soon.
Retentioneering 581 Jan 07, 2023
A Python package for modular causal inference analysis and model evaluations

Causal Inference 360 A Python package for inferring causal effects from observational data. Description Causal inference analysis enables estimating t

International Business Machines 506 Dec 19, 2022
Statsmodels: statistical modeling and econometrics in Python

About statsmodels statsmodels is a Python package that provides a complement to scipy for statistical computations including descriptive statistics an

statsmodels 8k Dec 29, 2022
Data and code accompanying the paper Politics and Virality in the Time of Twitter

Politics and Virality in the Time of Twitter Data and code accompanying the paper Politics and Virality in the Time of Twitter. In specific: the code

Cardiff NLP 3 Jul 02, 2022
This is a python script to navigate and extract the FSD50K dataset

FSD50K navigator This is a script I use to navigate the sound dataset from FSK50K.

sweemeng 2 Nov 23, 2021
A pipeline that creates consensus sequences from a Nanopore reads. I

A pipeline that creates consensus sequences from a Nanopore reads. It clusters reads that are similar to each other and creates a consensus that is then identified using BLAST.

Ada Madejska 2 May 15, 2022
A real-time financial data streaming pipeline and visualization platform using Apache Kafka, Cassandra, and Bokeh.

Realtime Financial Market Data Visualization and Analysis Introduction This repo shows my project about real-time stock data pipeline. All the code is

6 Sep 07, 2022
sportsdataverse python package

sportsdataverse-py See CHANGELOG.md for details. The goal of sportsdataverse-py is to provide the community with a python package for working with spo

Saiem Gilani 37 Dec 27, 2022
Zipline, a Pythonic Algorithmic Trading Library

Zipline is a Pythonic algorithmic trading library. It is an event-driven system for backtesting. Zipline is currently used in production as the backte

Quantopian, Inc. 15.7k Jan 07, 2023
Advanced Pandas Vault — Utilities, Functions and Snippets (by @firmai).

PandasVault ⁠— Advanced Pandas Functions and Code Snippets The only Pandas utility package you would ever need. It has no exotic external dependencies

Derek Snow 374 Jan 07, 2023
Jupyter notebooks for the book "The Elements of Statistical Learning".

This repository contains Jupyter notebooks implementing the algorithms found in the book and summary of the textbook.

Madiyar 369 Dec 30, 2022
Python package to transfer data in a fast, reliable, and packetized form.

pySerialTransfer Python package to transfer data in a fast, reliable, and packetized form.

PB2 101 Dec 07, 2022
2019 Data Science Bowl

Kaggle-2019-Data-Science-Bowl-Solution - Here i present my solution to kaggle 2019 data science bowl and how i improved it to win a silver medal in that competition.

Deepak Nandwani 1 Jan 01, 2022
Handle, manipulate, and convert data with units in Python

unyt A package for handling numpy arrays with units. Often writing code that deals with data that has units can be confusing. A function might return

The yt project 304 Jan 02, 2023
Using approximate bayesian posteriors in deep nets for active learning

Bayesian Active Learning (BaaL) BaaL is an active learning library developed at ElementAI. This repository contains techniques and reusable components

ElementAI 687 Dec 25, 2022
Fit models to your data in Python with Sherpa.

Table of Contents Sherpa License How To Install Sherpa Using Anaconda Using pip Building from source History Release History Sherpa Sherpa is a modeli

134 Jan 07, 2023
PyTorch implementation for NCL (Neighborhood-enrighed Contrastive Learning)

NCL (Neighborhood-enrighed Contrastive Learning) This is the official PyTorch implementation for the paper: Zihan Lin*, Changxin Tian*, Yupeng Hou* Wa

RUCAIBox 73 Jan 03, 2023
Code for the DH project "Dhimmis & Muslims – Analysing Multireligious Spaces in the Medieval Muslim World"

Damast This repository contains code developed for the digital humanities project "Dhimmis & Muslims – Analysing Multireligious Spaces in the Medieval

University of Stuttgart Visualization Research Center 2 Jul 01, 2022
DataPrep — The easiest way to prepare data in Python

DataPrep — The easiest way to prepare data in Python

SFU Database Group 1.5k Dec 27, 2022
An Aspiring Drop-In Replacement for NumPy at Scale

Legate NumPy is a Legate library that aims to provide a distributed and accelerated drop-in replacement for the NumPy API on top of the Legion runtime. Using Legate NumPy you do things like run the f

Legate 502 Jan 03, 2023