REVIEW 21 cited by
Lagrangian Neural Networks
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Accurate models of the world are built upon notions of its underlying symmetries. In physics, these symmetries correspond to conservation laws, such as for energy and momentum. Yet even though neural network models see increasing use in the physical sciences, they struggle to learn these symmetries. In this paper, we propose Lagrangian Neural Networks (LNNs), which can parameterize arbitrary Lagrangians using neural networks. In contrast to models that learn Hamiltonians, LNNs do not require canonical coordinates, and thus perform well in situations where canonical momenta are unknown or difficult to compute. Unlike previous approaches, our method does not restrict the functional form of learned energies and will produce energy-conserving models for a variety of tasks. We test our approach on a double pendulum and a relativistic particle, demonstrating energy conservation where a baseline approach incurs dissipation and modeling relativity without canonical coordinates where a Hamiltonian approach fails. Finally, we show how this model can be applied to graphs and continuous systems using a Lagrangian Graph Network, and demonstrate it on the 1D wave equation.
Forward citations
Cited by 21 Pith papers
-
GENERIC-FNO: Embedding Energy Conservation and Entropy Production into Fourier Neural Operators
GENERIC-FNO is the first neural operator to embed the complete GENERIC (metriplectic) structure directly in function space, enforcing energy conservation and entropy production exactly by construction across multiple ...
-
HyperODE: Zero-Shot Surrogate for Simulation and Inference of Dynamical Systems
A single 13k-parameter hypergraph surrogate trained on simple mass-conserving ODEs predicts calibrated trajectory quantiles and infers parameters across unseen compartmental families, system sizes, interaction orders,...
-
The Equilibrium Is the Initialization: Lazy Identity Collapse in Physics-Structured Deep Equilibrium Reasoning
In a port-Hamiltonian DEQ with learned initialization, the equilibrium equals the start to numerical precision and contributes +0.00 pp accuracy in 18 of 19 runs; a four-test diagnostic exposes the no-op.
-
CaLiSym: Learning Symplectic Dynamics of Real-World Systems through Structured Canonical Lifts
Lifting non-conservative, actuated, and contact-constrained robot dynamics into an exactly symplectic phase-space map yields state-of-the-art out-of-distribution autoregressive rollout error at low parameter and FLOP cost.
-
When is a System Discoverable from Data? Discovery Requires Chaos
Uniquely identifying an ODE from trajectory data depends on the trajectory filling enough of the state space: chaos on a high-dimensional attractor yields analytic discoverability, while first integrals preclude it.
-
Learning Hamiltonian Dynamics at Scale: A Differential-Geometric Approach
RO-HNN combines a geometrically-constrained symplectic autoencoder with a geometric Hamiltonian neural network to enable physically consistent predictions on high-dimensional systems via model-order reduction.
-
Latent Lie-Poisson Neural Networks (LLPNNs): Discovering the motion of Lie-Poisson systems through observable data and latent dynamics
LLPNNs recover latent Lie–Poisson momentum dynamics from observable configuration and velocity data by exploiting conserved spatial momentum and coadjoint reconstruction.
-
Implicit Machine Learning Force Fields Accelerate Molecular Dynamics Simulations
Replacing explicit neural network stacks with self-consistent fixed-point iterations, and warm-starting the solver across timesteps, gives 2-5x cheaper molecular dynamics force evaluation at matched accuracy.
-
Can AI Follow In Einstein's Footsteps?
AI for physics has moved from explicit equation discovery to black-box prediction, a trajectory the authors argue reverses the historical progression of human physics, leaving the invention of new mathematical framewo...
-
Quantum Port-Hamiltonian Neural Networks: Learning Conservative and Dissipative Dynamics via Measurement-Induced Nonlinearity
Q-pHNNs learn classical conservative and dissipative dynamics by mapping the port-Hamiltonian J matrix to unitary gates and the R matrix to mid-circuit measurement nonlinearity, enforcing structure by construction.
-
Gradient Networks for Universal Magnetic Modeling of Synchronous Machines
A gradient-network model trained on sparse flux-linkage/current data reproduces the saturable, angle-periodic magnetic maps of a 5.6-kW synchronous machine while enforcing reciprocity, convexity, and smoothness by con...
-
Structure-Preserving Digital Twins via Conditional Neural Whitney Forms
A transformer-based architecture learns a structure-preserving reduced finite element model, with conservation laws held exactly by the finite element exterior calculus construction, for data-calibrated real-time digi...
-
Discovering Interpretable Ordinary Differential Equations from Noisy Data
A pipeline that fits noisy data with a smooth general-solution function, differentiates it with B-splines, and uses SVD to recover a constant-coefficient linear ODE.
-
Data-free neural PDE solvers based on Graph Neural Networks and weak forms
A graph-neural-network PDE solver trained on the weak-form force residual — no simulation data — reports residual convergence below 1% on unseen load cases and one modified geometry, with residual-based test-time refinement.
-
Geometry-Conditioned Fourier Neural Operators for Cubic Nonlinear Schrodinger Dynamics on Periodic Domains
A geometry-conditioned FNO is trained on pseudospectral data to approximate the one-step operator for cubic NLS on 2D tori and reproduces distinct H²-norm growth on rational versus irrational aspect ratios.
-
A hierarchy of thermodynamics learning frameworks for inelastic constitutive modeling
With identical neural building blocks, dissipation-potential, GSM, and metriplectic models all learn accurate inelastic stress responses; performance differences are modest and dataset-dependent.
-
Data Readiness for Scientific AI at Scale
Scientific data can be graded on a five-level readiness scale crossed with five processing stages, yielding a maturity matrix for AI training at supercomputer scale.
-
AI4Research: A Survey of Artificial Intelligence for Scientific Research
A survey that organizes AI-for-research work into five tasks, comprehension, survey, discovery, writing, and peer review, and compiles associated tools and benchmarks.
-
Newtonian and Lagrangian Neural Networks: A Comparison Towards Efficient Inverse Dynamics Identification
When motor torques are estimated rather than measured, Newtonian neural networks beat Lagrangian networks on a six-axis industrial robot because Lagrangian networks do not directly model friction and dissipation.
-
Investigating Lagrangian Neural Networks for Infinite Horizon Planning in Quadrupedal Locomotion
In Isaac Gym simulation, Lagrangian-structured dynamics models learn faster and predict motion more accurately than a plain MLP, but the headline 10x and 2-10x numbers are not fully supported.
-
Symmetry-preserving neural networks in lattice field theories
Translation- and gauge-equivariant neural networks (L-CNNs) predict Wilson loops, topological charge, and flux observables with orders-of-magnitude lower error than symmetry-breaking baselines, and neural gradient flo...
Discussion (0). Sign in to comment.