SPLICE: Sparse Physics Library Identification and Compositional Expansion
Not on the map. Why?
Abstract (arXiv)
Neural operator splitting composes a library of neural operators, one per physical mechanism, into zero-shot simulators of new equations. Existing approaches, however, require these mechanisms and their labels to be specified in advance. We ask whether a pretrained library can instead be adapted and grown from unlabeled trajectories in which multiple mechanisms act simultaneously. We introduce SPLICE, a framework that discovers operators without mechanism labels on mixture data. The contributions of the candidate operators to each trajectory are inferred by solving a convex optimization problem. A calibrated test then assigns each discovered operator the mechanism it implements. Failure cases are also detectable, when the library lacks the required physics or when mechanisms are insufficiently separable. Across one and two dimensional systems, the discovered operators recover mechanisms from unlabeled mixtures and compose into zero-shot simulators of unseen equations; a frozen library can also grow an operator for physics absent from pretraining, without labels and without forgetting. On Navier-Stokes, the discovered operators compose about as accurately as their supervised counterparts.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 29 of its 43 references that are in the map. First posted on arXiv 12 Oct 2026.
Builds on
29 of 43 references.
References · 29
- 2018Neural Ordinary Differential Equationscited by 3,779
- 2020Fourier Neural Operator for Parametric Partial Differential Equationscited by 2,908
- 2019DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operatorscited by 2,018
- 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 941
- 2020Learning Mesh-Based Simulation with Graph Networkscited by 868
- 2021Machine learning accelerated computational fluid dynamicscited by 471
- 2022PDEBENCH: An Extensive Benchmark for Scientific Machine Learningcited by 413
- 2020Discovering Symbolic Models from Deep Learning with Inductive Biasescited by 344
- 2017PDE-Net: Learning PDEs from Datacited by 311
- 2021Modern Koopman Theory for Dynamical Systemscited by 276
- 2019Continual learning with hypernetworkscited by 273
- 2024Poseidon: Efficient Foundation Models for PDEscited by 200
17 more
- 2024DPOT: Auto-Regressive Denoising Operator Transformer for Large-Scale PDE Pre-Trainingcited by 151
- 2024The Well: a Large-Scale Collection of Diverse Physics Simulations for Machine Learningcited by 145
- 2021Ensemble-SINDy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and controlcited by 137
- 2017Learning Independent Causal Mechanismscited by 105
- 2023In-Context Operator Learning with Data Prompts for Differential Equation Problemscited by 101
- 2023Multiple Physics Pretraining for Physical Surrogate Modelscited by 98
- 2019DeepMoD: Deep learning for Model Discovery in noisy datacited by 71
- 2022Generalizing to New Physical Systems via Context-Informed Dynamics Modelcited by 50
- 2023Self-supervised learning of Split Invariant Equivariant representationscited by 44
- 2021LEADS: Learning Dynamical Systems that Generalize Across Environmentscited by 40
- 2021Meta-Learning Dynamics Forecasting Using Task Inferencecited by 38
- 2020Lifelong Learning of Compositional Structurescited by 35
- 2024Zebra: In-Context Generative Pretraining for Solving Parametric PDEscited by 16
- 2023A foundational neural operator that continuously learns without forgettingcited by 11
- 2026Test-time Generalization for Physics through Neural Operator Splittingcited by 10
- 2025Merging Memory and Space: A State Space Neural Operatorcited by 2
- 2026HyCOP: Hybrid Composition Operators for Interpretable Learning of PDEscited by 2
Built on by
Not cited yet.