Evaluating Graph Generative Models with Contrastively Learned Features (2022-06-13T00:00:00.000000Z)

TL;DR

It is demonstrated that Graph Substructure Networks (GSNs), which in a way combine both approaches, are better at distinguishing the distances between graph datasets.

Abstract

A wide range of models have been proposed for Graph Generative Models, necessitating effective methods to evaluate their quality. So far, most techniques use either traditional metrics based on subgraph counting, or the representations of randomly initialized Graph Neural Networks (GNNs). We propose using representations from contrastively trained GNNs, rather than random GNNs, and show this gives more reliable evaluation metrics. Neither traditional approaches nor GNN-based approaches dominate the other, however: we give examples of graphs that each approach is unable to distinguish. We demonstrate that Graph Substructure Networks (GSNs), which in a way combine both approaches, are better at distinguishing the distances between graph datasets.

Authors

Hamed Shirzad

1 papers

Kaveh Hassani

1 papers

Danica J. Sutherland

1 papers

TL;DR

Abstract

Authors

References58 items

Board

Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap

On Evaluation Metrics for Graph Generative Models

Learning Graph Augmentations to Learn Graph Representations

Cross-Domain Few-Shot Graph Classification

Graph Contrastive Learning Automated

Evaluation Metrics for Graph Generative Models: Problems, Pitfalls, and Practical Solutions

Graph Contrastive Learning with Adaptive Augmentation

Graph Contrastive Learning with Augmentations

A Systematic Survey on Deep Generative Models for Graph Generation

Scalable Deep Generative Modeling for Sparse Graphs

GCC: Graph Contrastive Coding for Graph Neural Network Pre-Training

Improving Graph Neural Network Expressivity via Subgraph Isomorphism Counting

Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning

Deep Graph Contrastive Representation Learning

Contrastive Multi-View Representation Learning on Graphs

Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere

Reliable Fidelity and Diversity Metrics for Generative Models

Learning Deep Kernels for Non-Parametric Two-Sample Tests

Graph Deconvolutional Generation

A Simple Framework for Contrastive Learning of Visual Representations

Can graph neural networks count substructures?

Efficient Graph Generation with Graph Recurrent Attention Networks

InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization

GraphNVP: An Invertible Flow Model for Generating Molecular Graphs

How Powerful are Graph Neural Networks?

Deep Graph Infomax

Learning deep representations by mutual information estimation and maximization

Assessing Generative Models via Precision and Recall

MolGAN: An implicit generative model for small molecular graphs

On gradient regularizers for MMD GANs

GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Models

Spectral Normalization for Generative Adversarial Networks

Learning Deep Generative Models of Graphs

Pros and Cons of GAN Evaluation Measures

Which Training Methods for GANs do actually Converge?

Demystifying MMD GANs

Graph Attention Networks

GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium

Stabilizing Training of Generative Adversarial Networks through Regularization

Neural Message Passing for Quantum Chemistry

Wasserstein GAN

Towards Principled Methods for Training Generative Adversarial Networks

Semi-Supervised Classification with Graph Convolutional Networks

Improved Techniques for Training GANs

Rethinking the Inception Architecture for Computer Vision

A note on the evaluation of generative models

ImageNet Large Scale Visual Recognition Challenge

ZINC: A Free Tool to Discover Chemistry for Biology

A Kernel Two-Sample Test

Collective Classification in Network Data

Distinguishing enzyme structures from non-enzymes without alignments.

On Self-Supervised Image Representations for GAN Evaluation

TD-GEN: Graph Generation With Tree Decomposition

Bootstrapped Representation Learning on Graphs

Collective Classi!cation in Network Data

On the evolution of random graphs

you used crowdsourcing or conducted research with human subjects

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names