Alvaro Sanchez-Gonzalez — accepted AI-conference papers

2023

Learning rigid dynamics with face interaction graph networks

Kelsey R Allen, Yulia Rubanova, Tatiana Lopez-Guevara, William F Whitney, Alvaro Sanchez-Gonzalez, Peter Battaglia +1

ICLR 2023top-25%

Simulating rigid collisions among arbitrary shapes is notoriously difficult due to complex geometry and the strong non-linearity of the interactions. While graph neural network (GNN)-based models are effective at learning to simulate complex physical dynamics, such as fluids, cloth and articulated b…

Cited by 34 Source ↗PDF ↗Scholar ↗

2023

Pre-training via Denoising for Molecular Property Prediction

Sheheryar Zaidi, Michael Schaarschmidt, James Martens, Hyunjik Kim, Yee Whye Teh, Alvaro Sanchez-Gonzalez +3

ICLR 2023top-25%

Many important problems involving molecular property prediction from 3D structures have limited data, posing a generalization challenge for neural networks. In this paper, we describe a pre-training technique based on denoising that achieves a new state-of-the-art in molecular property prediction by…

Cited by 145 Source ↗PDF ↗code ↗Scholar ↗

2022

Constraint-based graph network simulator

Yulia Rubanova, Alvaro Sanchez-Gonzalez, Tobias Pfaff, Peter Battaglia

ICML 2022spotlight

In the area of physical simulations, nearly all neural-network-based methods directly predict future states from the input states. However, many traditional simulation engines instead model the constraints of the system and select the state which satisfies them. Here we present a framework for const…

Cited by 38 Source ↗PDF ↗Scholar ↗

2022

Graph network simulators can learn discontinuous, rigid contact dynamics

Kelsey R Allen, Tatiana Lopez Guevara, Yulia Rubanova, Kim Stachenfeld, Alvaro Sanchez-Gonzalez, Peter Battaglia +1

CoRL 2022poster

Recent years have seen a rise in techniques for modeling discontinuous dynamics, such as rigid contact or switching motion modes, using deep learning. A common claim is that deep networks are incapable of accurately modeling rigid-body dynamics without explicit modules for handling contacts, due to…

Cited by 45 Source ↗Scholar ↗

2022

Inverse Design for Fluid-Structure Interactions using Graph Network Simulators

Kelsey R Allen, Tatiana Lopez-Guavara, Kim Stachenfeld, Alvaro Sanchez-Gonzalez, Peter Battaglia, Jessica B Hamrick +1

NeurIPS 2022accept

Designing physical artifacts that serve a purpose---such as tools and other functional structures---is central to engineering as well as everyday human behavior. Though automating design using machine learning has tremendous promise, existing methods are often limited by the task-dependent distribut…

Cited by 20 Source ↗PDF ↗Scholar ↗

2022

Learned Simulators for Turbulence

Kim Stachenfeld, Drummond Buschman Fielding, Dmitrii Kochkov, Miles Cranmer, Tobias Pfaff, Jonathan Godwin +4

ICLR 2022poster

Turbulence simulation with classical numerical solvers requires high-resolution grids to accurately resolve dynamics. Here we train learned simulators at low spatial and temporal resolutions to capture turbulent dynamics generated at high resolution. We show that our proposed model can simulate tur…

Cited by 47 Source ↗PDF ↗Scholar ↗

2022

Simple GNN Regularisation for 3D Molecular Property Prediction and Beyond

Jonathan Godwin, Michael Schaarschmidt, Alexander L Gaunt, Alvaro Sanchez-Gonzalez, Yulia Rubanova, Petar Veličković +2

ICLR 2022poster

In this paper we show that simple noisy regularisation can be an effective way to address oversmoothing. We first argue that regularisers ad-dressing oversmoothing should both penalise node latent similarity and encourage meaningful node representations. From this observation we derive “Noisy Nodes”…

Cited by 147 Source ↗PDF ↗Scholar ↗

2021

Learning Mesh-Based Simulation with Graph Networks

Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, Peter Battaglia

ICLR 2021spotlight

Mesh-based simulations are central to modeling complex physical systems in many disciplines across science and engineering. Mesh representations support powerful numerical integration methods and their resolution can be adapted to strike favorable trade-offs between accuracy and efficiency. However,…

Cited by 1101 Source ↗PDF ↗code ↗Scholar ↗

2020

Combining Q-Learning and Search with Amortized Value Estimates

Jessica B. Hamrick, Victor Bapst, Alvaro Sanchez-Gonzalez, Tobias Pfaff, Theophane Weber, Lars Buesing +1

ICLR 2020poster

We introduce "Search with Amortized Value Estimates" (SAVE), an approach for combining model-free Q-learning with model-based Monte-Carlo Tree Search (MCTS). In SAVE, a learned prior over state-action values is used to guide MCTS, which estimates an improved set of state-action values. The new Q-est…

Cited by 65 Source ↗Scholar ↗

2020

Learning to Simulate Complex Physics with Graph Networks

Alvaro Sanchez-Gonzalez, Jonathan Godwin, Tobias Pfaff, Rex Ying, Jure Leskovec, Peter Battaglia

ICML 2020poster

Here we present a machine learning framework and model implementation that can learn to simulate a wide variety of challenging physical domains, involving fluids, rigid solids, and deformable materials interacting with one another. Our framework—which we term "Graph Network-based Simulators" (GNS)—r…

Cited by 1440 Source ↗PDF ↗code ↗Scholar ↗

2019

CompILE: Compositional Imitation Learning and Execution

Thomas Kipf, Yujia Li, Hanjun Dai, Vinicius Zambaldi, Alvaro Sanchez-Gonzalez, Edward Grefenstette +2

ICML 2019oral

We introduce Compositional Imitation Learning and Execution (CompILE): a framework for learning reusable, variable-length segments of hierarchically-structured behavior from demonstration data. CompILE uses a novel unsupervised, fully-differentiable sequence segmentation module to learn latent encod…

Cited by 163 Source ↗PDF ↗code ↗Scholar ↗

2019

Structured agents for physical construction

Victor Bapst, Alvaro Sanchez-Gonzalez, Carl Doersch, Kimberly Stachenfeld, Pushmeet Kohli, Peter Battaglia +1

ICML 2019oral

Physical construction—the ability to compose objects, subject to physical dynamics, to serve some function—is fundamental to human intelligence. We introduce a suite of challenging physical construction tasks inspired by how children play with blocks, such as matching a target configuration, stackin…

Cited by 126 Source ↗PDF ↗Scholar ↗

2018

Graph Networks as Learnable Physics Engines for Inference and Control

Alvaro Sanchez-Gonzalez, Nicolas Heess, Jost Tobias Springenberg, Josh Merel, Martin Riedmiller, Raia Hadsell +1

ICML 2018oral

Understanding and interacting with everyday physical scenes requires rich knowledge about the structure of the world, represented either implicitly in a value or policy function, or explicitly in a transition model. Here we introduce a new class of learnable models–based on graph networks–which impl…

Cited by 794 Source ↗PDF ↗Scholar ↗