Relational inductive biases, deep learning, and graph networks
2018/06/04 by Peter W. Battaglia, Battaglia, Peter W., Jessica B. Hamrick +51 · 2 voices · 185 citations
Computer Science · Mathematics · #cs.AI #cs.LG #stat.ML
paper · pdf · doi:10.48550/arxiv.1806.01261
arxiv created 2018/10/17 · arxiv updated 2018/10/18
Abstract
Artificial intelligence (AI) has undergone a renaissance recently, making major progress in key domains such as vision, language, control, and decision-making. This has been due, in part, to cheap data and cheap compute resources, which have fit the natural strengths of deep learning. However, many defining characteristics of human intelligence, which developed under much different pressures, remain out of reach for current approaches. In particular, generalizing beyond one's experiences--a hallmark of human intelligence from infancy--remains a formidable challenge for modern AI. The following is part position paper, part review, and part unification. We argue that combinatorial generalization must be a top priority for AI to achieve human-like abilities, and that structured representations and computations are key to realizing this objective. Just as biology uses nature and nurture cooperatively, we reject the false choice between "hand-engineering" and "end-to-end" learning, and instead advocate for an approach which benefits from their complementary strengths. We explore how using relational inductive biases within deep learning architectures can facilitate learning about entities, relations, and rules for composing them. We present a new building block for the AI toolkit with a strong relational inductive bias--the graph network--which generalizes and extends various approaches for neural networks that operate on graphs, and provides a straightforward interface for manipulating structured knowledge and producing structured behaviors. We discuss how graph networks can support relational reasoning and combinatorial generalization, laying the foundation for more sophisticated, interpretable, and flexible patterns of reasoning. As a companion to this paper, we have released an open-source software library for building graph networks, with demonstrations of how to use them in practice.
Cited by
- LLM-Driven Cross-Paradigm Design for Quantum Optimal Control
- SPIN: Decentralized Swarm Control via Tensorized Policy Coordination
- Planning as Emergent Behavior in Reinforcement Learning with Relational Hidden States
- Spatio-Temporal Prediction of Unsteady Airfoil Aerodynamics Using Augmented Graph Neural Ordinary Differential Equations with Exogenous Controls
- Machine learning of electronic structure and atomistic properties from the external potential
- Pretrained Event Classification Model for High Energy Physics Analysis
- Fast Wasserstein rates for estimating probability distributions of probabilistic graphical models
- Toward manifest relationality in transformers via symmetry reduction
- Intelligent Three Level Learning Architecture for Autonomous UAV Swarms in Search and Rescue
- Beyond Statistical Learning: Exact Learning Is Essential for General Intelligence
- Transformers are Graph Neural Networks
- Relational reasoning and inductive bias in transformers and large language models
- Extending the RANGE of Graph Neural Networks: Relaying Attention Nodes for Global Encoding
- With Great Context Comes Great Prediction Power: Classifying Objects via Geo-Semantic Scene Graphs
- Debugging Tabular Log as Dynamic Graphs
- Learning to Access Computation: Accessibility Plasticity as a Principle of Adaptive Intelligence
- TSA: Temporal Slot Activation for Persistent Object-Centric Video Representation
- Sheet As Token: A Graph-Enhanced Representation for Multi-Sheet Spreadsheet Understanding
- Unifying Deep Predicate Invention with Pre-trained Foundation Models
- How Smoothing is N-simplicial Attention?
- Dual-Qubit Hierarchical Fuzzy Neural Network for Image Classification: Enabling Relational Learning via Quantum Entanglement
- Bridging Data and Physics: A Graph Neural Network-Based Hybrid Twin Framework
- An Operator-Consistent Graph Neural Network for Learning Diffusion Dynamics on Irregular Meshes
- Physics-Informed Learning of Microvascular Flow Models using Graph Neural Networks
- Supervised Learning of Random Neural Architectures Structured by Latent Random Fields on Compact Boundaryless Multiply-Connected Manifolds
- Graph Deep Learning for Intracranial Aneurysm Blood Flow Simulation and Risk Assessment
- Mitigating Bias in Graph Hyperdimensional Computing
- The Graph-Embedded Hazard Model (GEHM): Stochastic Network Survival Dynamics on Economic Graphs
- Learning the Cosmic Web: Graph-based Classification of Simulated Galaxies by their Dark Matter Environments
- What Happens When: Learning Temporal Orders of Events in Videos
- Representation of Inorganic Synthesis Reactions and Prediction: Graphical Framework and Datasets
- Graph Query Networks for Object Detection with Automotive Radar
- Warm-starting active-set solvers using graph neural networks
- A mathematical theory of balancing relational generalization and memorization
- Improving Long-Range Interactions in Graph Neural Simulators via Hamiltonian Dynamics
- Parameter-Efficient Conditioning for Material Generalization in Graph-Based Simulators
- Context-aware Learned Mesh-based Simulation via Trajectory-Level Meta-Learning
- Accelerating scientific discovery with the common task framework
- Causal Graph Neural Networks for Healthcare
- A Roadmap for Predictive Human Immunology
- Graph-Based Alternatives to LLMs for Human Simulation
- Solution Space Topology Guides CMTS Search
- Cloning Deterministic Worlds: The Critical Role of Latent Geometry in Long-Horizon World Models
- Graph Network-based Structural Simulator: Graph Neural Networks for Structural Dynamics
- Symmetry and Generalisation in Neural Approximations of Renormalisation Transformations
- Enhanced spatial clustering of single-molecule localizations with graph neural networks
- Exploring an image-based b-jet tagging method using convolution neural networks
- Logical GANs: Adversarial Learning through Ehrenfeucht Fraisse Games
- A Hybrid GNN-LSE Method for Fast, Robust, and Physically-Consistent AC Power Flow
- MEIcoder: Decoding Visual Stimuli from Neural Activity by Leveraging Most Exciting Inputs
- UniCode: Augmenting Evaluation for Code Reasoning
- AGNES: Adaptive Graph Neural Network and Dynamic Programming Hybrid Framework for Real-Time Nanopore Seed Chaining
- Multi-View Graph Feature Propagation for Privacy Preservation and Feature Sparsity
- D3MAS: Decompose, Deduce, and Distribute for Enhanced Knowledge Sharing in Multi-Agent Systems
- Color3D: Controllable and Consistent 3D Colorization with Personalized Colorizer
- Compositional meta-learning through probabilistic task inference
- Multi-View Graph Learning with Graph-Tuple
- MaNGO - Adaptable Graph Network Simulators via Meta-Learning
- Uncertainty assessment in satellite-based greenhouse gas emissions estimates using emulated atmospheric transport
- Convolution and Graph-based Deep Learning Approaches for Gamma/Hadron Separation in Imaging Atmospheric Cherenkov Telescopes
- When Does Global Attention Help? A Unified Empirical Study on Atomistic Graph Learning
- The Role of Feature Interactions in Graph-based Tabular Deep Learning
- Neural Message-Passing on Attention Graphs for Hallucination Detection
- Graph-Based Learning of Free Surface Dynamics in Generalized Newtonian Fluids using Smoothed Particle Hydrodynamics
- Aligning Inductive Bias for Data-Efficient Generalization in State Space Models
- Data-free neural PDE solvers based on Graph Neural Networks and weak forms
- Chem World: A Large-Scale Benchmark and Physics-Informed Framework for Trustworthy Chemical Property Prediction
- Schreier-Coset Graph Rewiring
- Study Design and Demystification of Physics Informed Neural Networks for Power Flow Simulation
- Fast Linear Solvers via AI-Tuned Markov Chain Monte Carlo-based Matrix Inversion
- On the Limits of Tabular Hardness Metrics for Deep RL: A Study with the Pharos Benchmark
- Deep Learning Inductive Biases for fMRI Time Series Classification during Resting-state and Movie-watching
- Automated Cyber Defense with Generalizable Graph-based Reinforcement Learning Agents
- SAMPO:Scale-wise Autoregression with Motion PrOmpt for generative world models
- AI-Driven Multi-Agent Vehicular Planning for Battery Efficiency and QoS in 6G Smart Cities
- Hierarchical Learning for Maze Navigation: Emergence of Mental Representations via Second-Order Learning
- A Scenario-Driven Cognitive Approach to Next-Generation AI Memory
- DeepTraverse: A Depth-First Search Inspired Network for Algorithmic Visual Understanding
- Score Matching on Large Geometric Graphs for Cosmology Generation
- How are Scientific Concepts Birthed? Typing Rules of Concept Formation in Theoretical Physics Reasoning
- A neural drift-plus-penalty algorithm for network power allocation and routing
- One Model, Two Minds: A Context-Gated Graph Learner that Recreates Human Biases
- Learning to accelerate distributed ADMM using graph neural networks
- RoboBallet: Planning for Multi-Robot Reaching with Graph Neural Networks and Reinforcement Learning
- Hot-Ham: an accurate and efficient E(3)-equivariant machine-learning electronic structures calculation framework
- Error analysis for learning the time-stepping operator of evolutionary PDEs
- Visible Yet Unreadable: A Systematic Blind Spot of Vision Language Models Across Writing Systems
- Learning Particle Dynamics Subject to Rigid Body Manipulations Using Graph Neural Networks
- Quantum latent distributions in deep generative models
- Automated discovery of finite volume schemes using Graph Neural Networks
- STDiff: A State Transition Diffusion Framework for Time Series Imputation in Industrial Systems
- On the Generalisation of Koopman Representations for Chaotic System Control
- From Prediction to Simulation: AlphaFold 3 as a Differentiable Framework for Structural Biology
- PKG-DPO: Optimizing Domain-Specific AI systems with Physics Knowledge Graphs and Direct Preference Optimization
- Training Transformers for Mesh-Based Simulations
- Graph Structure Learning with Temporal Graph Information Bottleneck for Inductive Representation Learning
- The Social Context of Human-Robot Interactions
- A Fully Spectral Neuro-Symbolic Reasoning Architecture with Graph Signal Processing as the Computational Backbone
- Scaling Up without Fading Out: Goal-Aware Sparse GNN for RL-based Generalized Planning
- Reasoning in Computer Vision: Taxonomy, Models, Tasks, and Methodologies
- Graph Neural Network Surrogates for Contacting Deformable Bodies with Necessary and Sufficient Contact Detection
- Artificial intelligence in mitotic checkpoint modeling: transforming our understanding of cellular division through machine learning and predictive biology
- Deeper Inside Deep ViT
- Forecasting NCAA Basketball Outcomes with Deep Learning: A Comparative Study of LSTM and Transformer Models
- Invariant Graph Transformer for Out-of-Distribution Generalization
- Hierarchical Message-Passing Policies for Multi-Agent Reinforcement Learning
- GraphTorque: Torque-Driven Rewiring Graph Neural Network
- Neural Preconditioning via Krylov Subspace Geometry
- Enhancing Spatial Reasoning through Visual and Textual Thinking
- StackTrans: From Large Language Model to Large Pushdown Automata Model
- Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning
- Learning Long-Range Representations with Equivariant Messages
- GIIFT: Graph-guided Inductive Image-free Multimodal Machine Translation
- Deep Learning for Blood-Brain Barrier Permeability Prediction: From Discriminative Models to Mechanism-Aware Design
- Resolving Indirect Calls in Binary Code via Cross-Reference Augmented Graph Neural Networks
- Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem
- Reconstruction of Dark Matter and Baryon Density From Galaxies: A Comparison of Linear, Halo Model and Machine Learning-Based Methods
- NeurIPS 2024 ML4CFD Competition: Results and Retrospective Analysis
- Machine Learning for the Cluster Reconstruction in the CALIFA Calorimeter at R3B
- Robust Noise Attenuation via Adaptive Pooling of Transformer Outputs
- Learning to Quantize and Precode in Massive MIMO Systems for Energy Reduction: a Graph Neural Network Approach
- A document is worth a structured record: Principled inductive bias design for document recognition
- Recurrent U-Net-Based Graph Neural Network (RUGNN) for Accurate Deformation Predictions in Sheet Material Forming
- Scaling can lead to compositional generalization
- MolSight: A Graph-Aware Vision-Language Model for Unified Chemical Image Understanding
- Relational inductive biases on attention mechanisms
- Direct Vertex Reconstruction of Λ Baryons from Hits in CLAS12 using Graph Neural Networks
- VOCAL: Visual Odometry via ContrAstive Learning
- When GNNs Met a Word Equations Solver: Learning to Rank Equations (Extended Technical Report)
- ParticleFormer: A 3D Point Cloud World Model for Multi-Object, Multi-Material Robotic Manipulation
- Identifiable Object Representations under Spatial Ambiguities
- DistShap: Scalable GNN Explanations with Distributed Shapley Values
- Active Inference AI Systems for Scientific Discovery
- Geometric deep learning assists protein engineering. Opportunities and Challenges
- Mitigating Over-Squashing in Graph Neural Networks by Spectrum-Preserving Sparsification
- Over-squashing in Spatiotemporal Graph Neural Networks
- Toward a Graph Foundation Model: Pre-Training Transformers With Random Walks
- Training Neural Networks by Optimizing Neuron Positions
- Learning geometric invariant features for classification of vector polygons with graph message-passing neural network
- CheMixHub: Datasets and Benchmarks for Chemical Mixture Property Prediction
- End-to-End Probabilistic Framework for Learning with Hard Constraints
- SIMSHIFT: A Benchmark for Adapting Neural Surrogates to Distribution Shifts
- FIGNN: Feature-Specific Interpretability for Graph Neural Network Surrogate Models
- Graph Semi-Supervised Learning for Point Classification on Data Manifolds
- LaDEEP: A Deep Learning-based Surrogate Model for Large Deformation of Elastic-Plastic Solids
- Flow-Attentional Graph Neural Networks
- Joint Beamforming and Integer User Association using a GNN with Gumbel-Softmax Reparameterizations
- Two-dimensional Taxonomy for N-ary Knowledge Representation Learning Methods
- Trajectory First: A Curriculum for Discovering Diverse Policies
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective
- Beyond Attention: Learning Spatio-Temporal Dynamics with Emergent Interpretable Topologies
- Deep Learning Weather Models for Subregional Ocean Forecasting: A Case Study on the Canary Current Upwelling System
- Simulating the Unseen: Crash Prediction Must Learn from What Did Not Happen
- Semantic Communication meets System 2 ML: How Abstraction, Compositionality and Emergent Languages Shape Intelligence
- Learning simple heuristic rules for classifying materials based on chemical composition
- Graph-Based Operator Learning from Limited Data on Irregular Domains
- Hierarchical-embedding autoencoder with a predictor (HEAP) as efficient architecture for learning long-term evolution of complex multi-scale physical systems
- How Particle System Theory Enhances Hypergraph Message Passing
- Few-Shot Learning from Gigapixel Images via Hierarchical Vision-Language Alignment and Modeling
- On the use of Graphs for Satellite Image Time Series
- Graph-Conditional Flow Matching for Relational Data Generation
- Lightweight and Interpretable Transformer via Mixed Graph Algorithm Unrolling for Traffic Forecast
- Multi-Level Monte Carlo Training of Neural Operators
- Unsupervised Learning of Local Updates for Maximum Independent Set in Dynamic Graphs
- Structured Relational Representations
- SAR-GTR: Attributed Scattering Information Guided SAR Graph Transformer Recognition Algorithm
- Numerical Solution of Mixed-Dimensional PDEs Using a Neural Preconditioner
- GraphComp: Extreme Error-bounded Compression of Scientific Data via Temporal Graph Autoencoders
- Graph-based Agent Memory: Taxonomy, Techniques, and Applications
- Holos: A Web-Scale LLM-Based Multi-Agent System for the Agentic Web
- One Scale at a Time: Scale-Autoregressive Modeling for Fluid Flow Distributions
- DOGMA: Weaving Structural Information into Data-centric Single-cell Transcriptomics Analysis
- Scalable Multi-Task Learning for Particle Collision Event Reconstruction with Heterogeneous Graph Neural Networks
- Neuro-Relational Programs: Unifying Queries and Neural Computation over Structured Data
- Flow-based Extremal Mathematical Structure Discovery
- Prototype-enhanced prediction in graph neural networks for climate applications
- PRIMAL3: Pathfinding via Reinforcement and Imitation Multi-Agent Learning - Leveraging LaCAM3
- Link prediction on multi-relational graphs from an influence propagation perspective
- FM4WiFi: Flow Matching for Multi-AP Coordination in Dense Deployments of Beyond Wi-Fi 8 Networks
- Differentiable graph neural network simulator for forward and inverse modeling of multi-layered slope system with multiple material properties
- Beamforming Design and Association Scheme for Multi-RIS Multi-User mmWave Systems Through Graph Neural Networks
- Visualization Tasks for Unlabelled Graphs
- Integrating Implicit and Explicit Relational Biases through Graph-Based Multiple Instance Learning: A Case Study in Skin Lesion Diagnosis
- Trajectory Encoding Temporal Graph Networks
- Prediction of Usage Probabilities of Shopping-Mall Corridors Using Heterogeneous Graph Neural Networks
Discussions
Related