Latent Dynamics · 2026-11-02

Meshless Latent Dynamics Models for Model Order Reduction of time-dependent Partial Differential Equations

Lorenzo Tomada, Federico Pichi, Gianluigi Rozza

arXiv:2610.41043PDFCode

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

The authors propose meshless Latent Dynamics Models for model order reduction of time-dependent, parameterized PDEs. They frame mesh-agnostic decoders as Neural Fields conditioned on a latent state, analyze expressivity–stability trade-offs via Lipschitz bounds for weak/strong/pure conditioning, and show grid-independent Lipschitz bounds in the meshless case. Experiments include Burgers’, Navier–Stokes flow past a cylinder, and linear advection.

Why it matters

Mesh-based latent dynamics tied to fixed discretizations cannot evaluate off-grid locations without interpolation errors or retraining. The paper’s unified conditioning framework and stability analysis target this limitation by providing theoretical guidance on how conditioning affects error amplification over time, and by quantifying discretization dependence through decoder Lipschitz constants. This supports transferable, mesh-agnostic MOR for time-dependent PDE families.

Method

  • Formulate mesh-agnostic LDMs by using Neural Fields (continuous spatial decoders) whose weights are conditioned on a latent state, covering weak, strong (FiLM), and pure conditioning mechanisms.
  • Derive a Universal Approximation Theorem (UAT) for weakly conditioned meshless LDMs and analyze stability via Lipschitz bounds for different conditioning mechanisms, yielding an expressivity–stability trade-off.
  • Compare meshless vs mesh-based architectures in the continuous mesh-refinement limit by analyzing Lipschitz constants, showing meshless models have grid-independent Lipschitz bounds; validate on Burgers’, Navier–Stokes, and linear advection.
Abstract (from arXiv)

Latent dynamics learning has emerged as a prominent approach for Model Order Reduction (MOR) of time-dependent, parameterized Partial Differential Equations (PDEs). Traditional autoencoder-based approaches are intrinsically tied to fixed spatial discretizations during both training and inference, motivating the development of meshless methodologies. In this context, conditioning provides a natural mechanism for representing temporal evolution through spatially continuous functions, enabling mesh-agnostic Latent Dynamics Models (LDMs). In this work, we introduce a unified mathematical framework for mesh-agnostic LDMs based on conditioning in Neural Fields (NeFs). Leveraging a taxonomy of conditioning mechanisms (weak, strong, and pure), we formulate meshless LDMs as NeFs conditioned on a latent state. We establish a Universal Approximation Theorem (UAT) for this class of models and analyze their stability by deriving Lipschitz bounds for different conditioning mechanisms. Our analysis reveals an expressivity--stability trade-off, providing insight into why more expressive conditioning, despite its higher representational capacity, can potentially degrade performance during temporal extrapolation. Finally, we provide a comprehensive comparison of meshless and mesh-based architectures in the continuous limit of the spatial discretization, showing that meshless models guarantee grid-independent Lipschitz bounds, unlike mesh-based counterparts. Our experiments across Burgers' equation, Navier--Stokes flow past a cylinder, and linear advection show that meshless LDMs can achieve predictive accuracy comparable to or better than their mesh-based counterparts while using only 2% of the trainable parameters and exhibiting nearly constant error across spatial discretizations.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: Date not supplied · Indexed by Pub-AI: 2026-11-02 UTC

preprint

Meshless Latent Dynamics Models for Model Order Reduction of time-dependent Partial Differential Equations

Lorenzo Tomada, Federico Pichi, Gianluigi Rozza

Source date: Date not supplied · Original record ↗

arxiv
2610.41043

    Imported from the existing registry. No new publication milestone asserted.

    Abstract · arxiv

    Latent dynamics learning has emerged as a prominent approach for Model Order Reduction (MOR) of time-dependent, parameterized Partial Differential Equations (PDEs). Traditional autoencoder-based approaches are intrinsically tied to fixed spatial discretizations during both training and inference, motivating the development of meshless methodologies. In this context, conditioning provides a natural mechanism for representing temporal evolution through spatially continuous functions, enabling mesh-agnostic Latent Dynamics Models (LDMs). In this work, we introduce a unified mathematical framework for mesh-agnostic LDMs based on conditioning in Neural Fields (NeFs). Leveraging a taxonomy of conditioning mechanisms (weak, strong, and pure), we formulate meshless LDMs as NeFs conditioned on a latent state. We establish a Universal Approximation Theorem (UAT) for this class of models and analyze their stability by deriving Lipschitz bounds for different conditioning mechanisms. Our analysis reveals an expressivity--stability trade-off, providing insight into why more expressive conditioning, despite its higher representational capacity, can potentially degrade performance during temporal extrapolation. Finally, we provide a comprehensive comparison of meshless and mesh-based architectures in the continuous limit of the spatial discretization, showing that meshless models guarantee grid-independent Lipschitz bounds, unlike mesh-based counterparts. Our experiments across Burgers' equation, Navier--Stokes flow past a cylinder, and linear advection show that meshless LDMs can achieve predictive accuracy comparable to or better than their mesh-based counterparts while using only 2% of the trainable parameters and exhibiting nearly constant error across spatial discretizations.