ICLR'22 | 图机器学习最近都在研究什么？

会员服务 ·

ICLR'22 | 图机器学习最近都在研究什么？

2021 年 11 月 19 日 图与推荐

ICLR由Yann LeCun和Yoshua Bengio等大牛发起，在2013年成立了第一届会议，会议开创了公开评议机制（open review）。它虽然是一个很年轻的会议，但已经成为深度学习领域不容忽视的顶级会议之一。

图机器学习包括图神经网络的很多论文都发表在ICLR上，例如17ICLR的GCN，18ICLR的GAT，19ICLR的PPNP等等。

关注了一波ICLR'22的投稿后，发现了一些图机器学习的热门研究方向，包括大规模GNN的scalability问题，深度GNN的过平滑问题，GNN的可解释性，自监督GNN等等热门研究方向。

ICLR 2022网址：https://openreview.net/group?id=ICLR.cc/2022/Conference

1.Scalability

Sampling Before Training: Rethinking the Effect of Edges in the Process of Training Graph Neural Networks
SpSC: A Fast and Provable Algorithm for Sampling-Based GNN Training
Revisiting Layer-wise Sampling in Fast Training for Graph Convolutional Networks
Large-Scale Representation Learning on Graphs via Bootstrapping
EXACT: Scalable Graph Neural Networks Training via Extreme Activation Compression
PipeGCN: Efficient Full-Graph Training of Graph Convolutional Networks with Pipelined Feature Communication
Increase and Conquer: Training Graph Neural Networks on Growing Graphs
LSP : Acceleration and Regularization of Graph Neural Networks via Locality Sensitive Pruning of Graphs
Embedding Compression with Hashing for Efficient Representation Learning in Graph
Locality-Based Mini Batching for Graph Neural Networks
Coarformer: Transformer for large graph via graph coarsening
Inductive Lottery Ticket Learning for Graph Neural Networks
IGLU: Efficient GCN Training via Lazy Updates
Graph Attention Multi-layer Perceptron
Learn Locally, Correct Globally: A Distributed Algorithm for Training Graph Neural Networks
Full-Precision Free Binary Graph Neural Networks
Graph-less Neural Networks: Teaching Old MLPs New Tricks Via Distillation
D2-GCN: Data-Dependent GCNs for Boosting Both Efficiency and Scalability
Adaptive Filters for Low-Latency and Memory-Efficient Graph Neural Networks

2.Oversmoothing/Depth

DeeperGCN: All You Need to Train Deeper GCNs
Understanding Graph Learning with Local Intrinsic Dimensionality
RankedDrop: Enhancing Deep Graph Convolutional Networks Training
DEEP GRAPH TREE NETWORKS
Evaluating Deep Graph Neural Networks
How Frequency Effect Graph Neural Networks
Revisiting Over-smoothing in BERT from the Perspective of Graph
Towards Feature Overcorrelation in Deeper Graph Neural Networks
Tackling Oversmoothing of GNNs with Contrastive Learning
Implicit vs Unfolded Graph Neural Networks
Towards Deepening Graph Neural Networks: A GNTK-based Optimization Perspective

3.Explainability

Explainability in Graph Convolutional Network for Representation Learning
FlowX: Towards Explainable Graph Neural Networks via Message Flows
On Theoretically and Empirically Analyzing GNN Explanation Methods
Task-Agnostic Graph Neural Explanations
Deconfounding to Explanation Evaluation in Graph Neural Networks
DEGREE: Decomposition Based Explanation for Graph Neural Networks
Discovering Invariant Rationales for Graph Neural Networks
Interpreting Graph Neural Networks via Unrevealed Causal Learning
Explainable GNN-Based Models over Knowledge Graphs

4.Self-Supervision

ESCo: Towards Provably Effective and Scalable Contrastive Representation Learning
m-mix: Generating hard negatives via multiple samples mixing for contrastive learning
Rethinking Temperature in Graph Contrastive Learning
Graph Barlow Twins: A self-supervised representation learning framework for graphs
Automated Self-Supervised Learning for Graphs
Self-Supervised Representation Learning via Latent Graph Prediction
Interrogating Paradigms in Self-supervised Graph Representation Learning
Self-Supervised Graph Neural Networks for Improved Electroencephalographic Seizure Analysis
Node Feature Extraction by Self-Supervised Multi-scale Neighborhood Prediction
SS-MAIL: Self-Supervised Multi-Agent Imitation Learning
Learning Graph Augmentations to Learn Graph Representations
Robust Graph Data Learning with Latent Graph Convolutional Representation

5. Adversarial Attacks / Robustness

GARNET: A Spectral Approach to Robust and Scalable Graph Neural Networks
Defending Graph Neural Networks via Tensor-Based Robust Graph Aggregation
A General Unified Graph Neural Network Framework Against Adversarial Attacks
Beyond Message Passing Paradigm: Training Graph Data with Consistency Constraints
Adversarial Weight Perturbation Improves Generalization in Graph Neural Networks
Bandits for Black-box Attacks to Graph Neural Networks with Structure Perturbation
Understanding and Improving Graph Injection Attack by Promoting Unnoticeability

6.Heterophily

On the Relationship between Heterophily and Robustness of Graph Neural Networks
Two Sides of the Same Coin: Heterophily and Oversmoothing in Graph Convolutional Neural Networks
An Interpretable Graph Generative Model with Heterophily
Is Heterophily A Real Nightmare For Graph Neural Networks on Performing Node Classification?
Is Homophily a Necessity for Graph Neural Networks?
Graph Information Matters: Understanding Graph Filters from Interaction Probability
GCN-SL: Graph Convolutional Network with Structure Learning for Disassortative Graphs

7. Heterogeneous Graphs

Equivariant Heterogeneous Graph Networks
Molecular Graph Representation Learning via Heterogeneous Motif Graph Construction
R-GSN: The Relation-based Graph Similar Network for Heterogeneous Graph

8. Multi-Relational Graphs

NodePiece: Compositional and Parameter-Efficient Representations of Large Knowledge Graphs
Neural Methods for Logical Reasoning over Knowledge Graphs
Time-aware Relational Graph Attention Network for Temporal Knowledge Graph Embeddings
A Topological View of Rule Learning in Knowledge Graphs
CareGraph: A Graph-based Recommender System for Diabetes Self-Care
Relational Multi-Task Learning: Modeling Relations between Data and Tasks
Inductive Relation Prediction Using Analogy Subgraph Embeddings

9. Hyper-relational Knowledge Graphs

Message Function Search for Hyper-relational Knowledge Graph
Query Embedding on Hyper-Relational Knowledge Graphs

10. Hypergraphs

You are AllSet: A Multiset Function Framework for Hypergraph Neural Networks
Hypergraph Convolutional Networks via Equivalency between Hypergraphs and Undirected Graphs
On the Expressiveness and Learning of Relational Neural Networks on Hypergraphs
A molecular hypergraph convolutional network with functional group information
Efficient Training and Inference of Hypergraph Reasoning Networks
FEATURE-AUGMENTED HYPERGRAPH NEURAL NETWORKS
GENERALIZING LINK PREDICTION FOR HYPERGRAPHS

11. Link Prediction

Revisiting Virtual Nodes in Graph Neural Networks for Link Prediction
Counterfactual Graph Learning for Link Prediction
Benchmarking Graph Neural Networks on Dynamic Link Prediction
Few-shot graph link prediction with domain adaptation
Neural Link Prediction with Walk Pooling

12. Graph Classification

GiG: Graph in Graph, a model boosting graph classification and representation learning
G-Mixup: Graph Augmentation for Graph Classification
Structural Optimization Makes Graph Classification Simpler and Better
Intrusion-Free Graph Mixup
The Infinite Contextual Graph Markov Model
Geometric Random Walk Graph Neural Networks via Implicit Layers
Wasserstein Weisfeiler-Lehman Subtree Distance for Graph-Structured Data
Adaptive Graph Capsule Convolutional Networks

13. Expressivity

Expressiveness and Approximation Properties of Graph Neural Networks
Counting Substructures with Higher-Order Graph Neural Networks: Possibility and Impossibility Results
A New Perspective on "How Graph Neural Networks Go Beyond Weisfeiler-Lehman?"
Equivariant and Stable Positional Encoding for More Powerful Graph Neural Networks
On Locality in Graph Learning via Graph Neural Network
On the Effect of Input Perturbations for Graph Neural Networks
Local Permutation Equivariance For Graph Neural Networks
How Attentive are Graph Attention Networks?

14. Subgraphs

Equivariant Subgraph Aggregation Networks
Count-GNN: Graph Neural Networks for Subgraph Isomorphism Counting
From Stars to Subgraphs: Uplifting Any GNN with Local Structure Awareness
NeuroSED: Learning Subgraph Similarity via Graph Neural Networks
Learning Representations of Partial Subgraphs by Subgraph InfoMax
GLASS: GNN with Labeling Tricks for Subgraph Representation Learning

15. Equivariance

Frame Averaging for Invariant and Equivariant Network Design
Geometric and Physical Quantities improve E(3) Equivariant Message Passing
Symmetry-driven graph neural networks

16. Generalisability

Towards Distribution Shift of Node-Level Prediction on Graphs: An Invariance Perspective
A Closer Look at Distribution Shifts and Out-of-Distribution Generalization on Graphs
Learning Two-Step Hybrid Policy for Graph-Based Interpretable Reinforcement Learning

17. Graph Generative Models: Evaluation Metrics

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

18. Proteins

Fast fixed-backbone protein sequence and rotamer design
Independent SE(3)-Equivariant Models for End-to-End Rigid Protein Docking
De novo design of protein target specific scaffold-based Inhibitors via Reinforcement Learning
An Effective GCN-based Hierarchical Multi-label classification for Protein Function Prediction
Geometric Transformers for Protein Interface Contact Prediction
Iterative Refinement Graph Neural Network for Antibody Sequence-Structure Co-design
Granger causal inference on DAGs identifies genomic loci regulating transcription

19. Molecules

Spanning Tree-based Graph Generation for Molecules
Molecular Graph Generation via Geometric Scattering
Generating Realistic 3D Molecules with an Equivariant Conditional Likelihood Model
Chemical-Reaction-Aware Molecule Representation Learning
Relative Molecule Self-Attention Transformer
Differentiable Scaffolding Tree for Molecule Optimization
MS2-Transformer: An End-to-End Model for MS/MS-assisted Molecule Identification
Graph Piece: Efficiently Generating High-Quality Molecular Graphs with Substructures
Pre-training Molecular Graph Representation with 3D Geometry
Spherical Message Passing for 3D Molecular Graphs
GraphEBM: Towards Permutation Invariant and Multi-Objective Molecular Graph Generation
Data-Efficient Graph Grammar Learning for Molecular Generation
Learning to Extend Molecular Scaffolds with Structural Motifs
Stepping Back to SMILES Transformers for Fast Molecular Representation Inference
An Autoregressive Flow Model for 3D Molecular Geometry Generation from Scratch
Learning 3D Representations of Molecular Chirality with Invariance to Bond Rotations
GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation
3D-Transformer: Molecular Representation with Transformer in 3D Space
Crystal Diffusion Variational Autoencoder for Periodic Material Generation
Knowledge Guided Geometric Editing for Unsupervised Drug Design
Spatial Graph Attention and Curiosity-driven Policy for Antiviral Drug Discovery
MoReL: Multi-omics Relational Learning

20. Molecular Property Prediction

3D Pre-training improves GNNs for Molecular Property Prediction
Simple GNN Regularisation for 3D Molecular Property Prediction and Beyond

21. Retrosynthesis

SemiRetro: Semi-template framework boosts deep retrosynthesis prediction
Permutation invariant graph-to-sequence model for template-free retrosynthesis and reaction prediction

22. Time Series

Evaluating the Robustness of Time Series Anomaly and Intrusion Detection Methods against Adversarial Attacks
Spatiotemporal Representation Learning on Time Series with Dynamic Graph ODEs
Multivariate Time Series Forecasting with Latent Graph Inference
TAMP-S2GCNets: Coupling Time-Aware Multipersistence Knowledge Representation with Spatio-Supra Graph Convolutional Networks for Time-Series Forecasting
Neural graphical modelling in continuous-time: consistency guarantees and algorithms
Filling the G_ap_s: Multivariate Time Series Imputation by Graph Neural Networks
Graph-Guided Network for Irregularly Sampled Multivariate Time Series
GAETS: A Graph Autoencoder Time Series Approach Towards Battery Parameter Estimation
Graph-Augmented Normalizing Flows for Anomaly Detection of Multiple Time Series
Back2Future: Leveraging Backfill Dynamics for Improving Real-time Predictions in Future
Causal discovery from conditionally stationary time-series

23. PDE

Message Passing Neural PDE Solvers

Learning Time-dependent PDE Solver using Message Passing Graph Neural Networks
Learning to Solve PDE-constrained Inverse Problems with Graph Networks

24. Physics

SiT: Simulation Transformer for Particle-based Physics Simulation
Predicting Physics in Mesh-reduced Space with Temporal Attention
KINet: Keypoint Interaction Networks for Unsupervised Forward Modeling
Constraint-based graph network simulator
Boundary Graph Neural Networks for 3D Simulations
Constrained Graph Mechanics Networks
Ab-Initio Potential Energy Surfaces by Pairing GNNs with Neural Wave Functions
Learning the Dynamics of Physical Systems from Sparse Observations with Finite Element Networks

25. Dynamic / Temporal Graphs

Dynamic Graph Representation Learning via Graph Transformer Networks
Metric Learning on Temporal Graphs via Few-Shot Examples
Online graph nets
Space-Time Graph Neural Networks
Convolutional Neural Network Dynamics: A Graph Perspective

26. Traffic

Simpler can be better: Multi-level Abstraction with Graph Convolution Recurrent Neural Network cells for Traffic Prediction
Learning to Remember Patterns: Pattern Matching Memory Networks for Traffic Forecasting
A multi-domain splitting framework for time-varying graph structure

27. Combinatorial Optimisation

Learning to Solve Combinatorial Problems via Efficient Exploration
Graph Neural Network Guided Local Search for the Traveling Salesperson Problem
Edge Rewiring Goes Neural: Boosting Network Resilience via Policy Gradient
Learning to Solve an Order Fulfillment Problem in Milliseconds with Edge-Feature-Embedded Graph Attention
Neural Models for Output-Space Invariance in Combinatorial Problems
What’s Wrong with Deep Learning in Tree Search for Combinatorial Optimization
Mind Your Solver! On Adversarial Attack and Defense for Combinatorial Optimization

28. Natural Language Processing

Crossformer: Transformer with Alternated Cross-Layer Guidance
Constituency Tree Representation for Argument Unit Recognition
Learning Object-Oriented Dynamics for Planning from Text
Fact-driven Logical Reasoning- Fact-driven Logical Reasoning

29. Language Modelling

GNN-LM: Language Modeling based on Global Contexts via GNN
Understanding Knowledge Integration in Language Models with Graph Convolutions
GreaseLM: Graph REASoning Enhanced Language Models

30. Question Answering

KG-FiD: Infusing Knowledge Graph in Fusion-in-Decoder for Open-Domain Question Answering
GNN is a Counter? Revisiting GNN for Question Answering

31. Computer Vision

Graph Similarities and Dual Approach for Sequential Text-to-Image Retrieval
Breaking Down Questions for Outside-Knowledge VQA
Towards Generic Interface for Human-Neural Network Knowledge Exchange
Revisiting Skeleton-based Action Recognition
Unified Recurrence Modeling for Video Action Anticipation
From Graph Local Embedding to Deep Metric Learning
Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space
SSR-GNNs: Stroke-based Sketch Representation with Graph Neural Networks
Knowledge-driven Scene Priors for Semantic Audio-Visual Embodied Navigation

32. Point Clouds

Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework
Concentric Spherical GNN for 3D Representation Learning

33. Fairness

Fair Node Representation Learning via Adaptive Data Augmentation
Generalized Demographic Parity for Group Fairness

34. Privacy/Federated Learning

Node-Level Differentially Private Graph Neural Networks
Federated Learning with Heterogeneous Architectures using Graph HyperNetworks
Federated Inference through Aligning Local Representations and Learning a Consensus Graph

35. Programming

GRAPHIX: A Pre-trained Graph Edit Model for Automated Program Repair
Exploring General Intelligence of Program Analysis for Multiple Tasks
ReGVD: Revisiting Graph Neural Networks for Vulnerability Detection

36. Multi-Agent Learning

Transform2Act: Learning a Transform-and-Control Policy for Efficient Agent Design
LPMARL: Linear Programming based Implicit Task Assigment for Hiearchical Multi-Agent Reinforcement Learning
ScheduleNet: Learn to solve multi-agent scheduling problems with reinforcement learning
Help Me Explore: Minimal Social Interventions for Graph-Based Autotelic Agents
A Multi-Agent Koopman Operator Approach for Distributed Representation Learning of Networked Dynamical Systems
Multi-Agent Decentralized Belief Propagation on Graphs
Context-Aware Sparse Deep Coordination Graphs
Learning to Solve Multi-Robot Task Allocation with a Covariant-Attention based Neural Architecture

37. Stochastic Block Models

LEARNING GUARANTEES FOR GRAPH CONVOLUTIONAL NETWORKS ON THE STOCHASTIC BLOCK MODEL
DIGRAC: Digraph Clustering Based on Flow Imbalance

38. Architecture Search

AutoCoG: A Unified Data-Modal Co-Search Framework for Graph Neural Networks
A Transferable General-Purpose Predictor for Neural Architecture Search

39. Healthcare

Using Graph Representation Learning with Schema Encoders to Measure the Severity of Depressive Symptoms
Deep Representations for Time-varying Brain Datasets
A Biologically Interpretable Graph Convolutional Network to Link Genetic Risk Pathways and Imaging Phenotypes of Disease

40. Miscellaneous

Neural Structured Prediction for Inductive Node Classification
Embedding models through the lens of Stable Coloring
GraphENS: Neighbor-Aware Ego Network Synthesis for Class-Imbalanced Node Classification
PI-GNN: Towards Robust Semi-Supervised Node Classification against Noisy Labels
PERSONALIZED LAB TEST RESPONSE PREDICTION WITH KNOWLEDGE AUGMENTATION
Learning Scenario Representation for Solving Two-stage Stochastic Integer Programs
WeaveNet: A Differentiable Solver for Non-linear Assignment Problems
Factored World Models for Zero-Shot Generalization in Robotic Manipulation
On Exploring Node-feature and Graph-structure Diversities for Node Drop Graph Pooling
Learning Graph Structure from Convolutional Mixtures
Large-Scale Adversarial Attacks on Graph Neural Networks via Graph Coarsening
Graph Condensation for Graph Neural Networks
Graph Kernel Neural Networks
Multiresolution Equivariant Graph Variational Autoencoder
Backpropagation-free Graph Convolutional Networks
Graph Neural Networks with Learnable Structural and Positional Representations
NAFS: A Simple yet Tough-to-Beat Baseline for Graph Representation Learning
SpecTRA: Spectral Transformer for Graph Representation Learning
A Deep Latent Space Model for Directed Graph Representation Learning
Convergent Graph Solvers
Effective Polynomial Filter Adaptation for Graph Neural Networks
G3: Representation Learning and Generation for Geometric Graphs
On the Unreasonable Effectiveness of Feature Propagation in Learning on Graphs with Missing Node Features
PF-GNN: Differentiable particle filtering based approximation of universal graph representations
Learning Graph Representations for Influence Maximization
Connecting Graph Convolution and Graph PCA
p-Laplacian Based Graph Neural Networks
PACE: A Parallelizable Computation Encoder for Directed Acyclic Graphs
Topological Graph Neural Networks
Towards Training Billion Parameter Graph Neural Networks for Atomic Simulations
Top-N: Equivariant Set and Graph Generation without Exchangeability
An Analysis of Attentive Walk-Aggregating Graph Neural Networks
Input Convex Graph Neural Networks: An Application to Optimal Control and Design Optimization
Spiking Graph Convolutional Networks
Trading Quality for Efficiency of Graph Partitioning: An Inductive Method across Graphs
Graph Auto-Encoder via Neighborhood Wasserstein Reconstruction
Personalized PageRank meets Graph Attention Networks
Understanding over-squashing and bottlenecks on graphs via curvature
Weakly Supervised Graph Clustering
Graph Tree Neural Networks
Open Set Domain Adaptation with Zero-shot Learning on Graph
Graph Convolutional Memory using Topological Priors
Learning to Pool in Graph Neural Networks for Extrapolation
Edge Partition Modulated Graph Convolutional Networks
Local Augmentation for Graph Neural Networks
GRAND++: Graph Neural Diffusion with A Source Term
Learning to Schedule Learning rate with Graph Neural Networks
Genome Sequence Reconstruction Using Gated Graph Convolutional Network
Graph Convolutional Networks via Adaptive Filter Banks
SONG: Self-Organizing Neural Graphs
Information Gain Propagation: a New Way to Graph Active Learning with Soft Labels
Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods
GIR Framework: Learning Graph Positional Embeddings with Anchor Indication and Path Encoding
Why Propagate Alone? Parallel Use of Labels and Features on Graphs
Stabilized Self-training with Negative Sampling on Few-labeled Graph Data
Accelerating Optimization using Neural Reparametrization
Convergence of Generalized Belief Propagation Algorithm on Graphs with Motifs
Convolutional Networks on Enhanced Message-Passing Graph Improve Semi-Supervised Classification with Few Labels
Efficient Ensembles of Graph Neural Networks
Learning to Infer the Structure of Network Games
Convergent Boosted Smoothing for Modeling GraphData with Tabular Node Features
Neural Relational Inference with Node-Specific Information
Triangle and Four Cycle Counting with Predictions in Graph Streams
Neurally boosted supervised spectral clustering
Structure-Aware Transformer Policy for Inhomogeneous Multi-Task Reinforcement Learning
Reasoning-Modulated Representations
Know Your Action Set: Learning Action Relations for Reinforcement Learning
Second-Order Unsupervised Feature Selection via Knowledge Contrastive Distillation