Optimal-Transport-Guided Functional Flow Matching for Turbulent Field Generation in Hilbert Space
arXiv:2604.05700v1 Announce Type: new Abstract: High-fidelity modeling of turbulent flows requires capturing complex spatiotemporal dynamics and multi-scale intermittency, posing a fundamental challenge for traditional knowledge-based systems. While deep generative models, such as diffusion models and Flow Matching, have shown promising performance, they are fundamentally constrained by their discrete, pixel-based nature. This limitation restricts their applicability in turbulence computing, where data inherently exists in a functional form. To address this gap, we propose Functional Optimal Transport Conditional Flow Matching (FOT-CFM), a generative framework defined directly in infinite-dimensional function space. Unlike conventional approaches defined on fixed grids, FOT-CFM treats physical fields as elements of an infinite-dimensional Hilbert space, and learns resolution-invariant generative dynamics directly at the level of probability measures. By integrating Optimal Transport (
arXiv:2604.05700v1 Announce Type: new Abstract: High-fidelity modeling of turbulent flows requires capturing complex spatiotemporal dynamics and multi-scale intermittency, posing a fundamental challenge for traditional knowledge-based systems. While deep generative models, such as diffusion models and Flow Matching, have shown promising performance, they are fundamentally constrained by their discrete, pixel-based nature. This limitation restricts their applicability in turbulence computing, where data inherently exists in a functional form. To address this gap, we propose Functional Optimal Transport Conditional Flow Matching (FOT-CFM), a generative framework defined directly in infinite-dimensional function space. Unlike conventional approaches defined on fixed grids, FOT-CFM treats physical fields as elements of an infinite-dimensional Hilbert space, and learns resolution-invariant generative dynamics directly at the level of probability measures. By integrating Optimal Transport (OT) theory, we construct deterministic, straight-line probability paths between noise and data measures in Hilbert space. This formulation enables simulation-free training and significantly accelerates the sampling process. We rigorously evaluate the proposed system on a diverse suite of chaotic dynamical systems, including the Navier-Stokes equations, Kolmogorov Flow, and Hasegawa-Wakatani equations, all of which exhibit rich multi-scale turbulent structures. Experimental results demonstrate that FOT-CFM achieves superior fidelity in reproducing high-order turbulent statistics and energy spectra compared to state-of-the-art baselines.
Executive Summary
The article introduces Functional Optimal Transport Conditional Flow Matching (FOT-CFM), a novel generative modeling framework that addresses limitations in high-fidelity turbulent flow simulation by operating directly in infinite-dimensional Hilbert space. Unlike traditional discrete, pixel-based approaches, FOT-CFM treats physical fields as functional data, enabling resolution-invariant generative dynamics. By integrating optimal transport theory, the method constructs deterministic, straight-line probability paths between noise and data measures, facilitating simulation-free training and accelerated sampling. The framework is rigorously evaluated on chaotic dynamical systems, including Navier-Stokes and Kolmogorov Flow, demonstrating superior performance in reproducing turbulent statistics and energy spectra compared to state-of-the-art baselines.
Key Points
- ▸ FOT-CFM operates in infinite-dimensional Hilbert space, overcoming the constraints of discrete, grid-based generative models.
- ▸ The framework leverages optimal transport theory to construct deterministic, straight-line probability paths, enabling efficient simulation-free training and accelerated sampling.
- ▸ The approach achieves superior fidelity in reproducing high-order turbulent statistics and energy spectra across diverse chaotic dynamical systems, including Navier-Stokes and Kolmogorov Flow.
Merits
Theoretical Rigor
The integration of optimal transport theory with functional flow matching in Hilbert space provides a mathematically robust foundation, addressing gaps in conventional generative modeling for turbulent flows.
Computational Efficiency
The deterministic, straight-line probability paths enable simulation-free training and significantly accelerated sampling, improving computational feasibility for large-scale turbulent simulations.
Generality and Scalability
The resolution-invariant nature of FOT-CFM allows for scalable applications across diverse chaotic systems, demonstrating broad applicability beyond traditional turbulence modeling.
Demerits
Complexity of Implementation
The requirement to operate in infinite-dimensional Hilbert space introduces significant computational and implementation challenges, potentially limiting accessibility for practitioners without advanced mathematical training.
Data Requirements
The method's reliance on high-fidelity functional data for training may necessitate extensive and computationally expensive datasets, posing practical barriers in resource-constrained environments.
Validation Challenges
While the results are promising, the validation of resolution-invariant generative dynamics in real-world scenarios beyond benchmark systems remains an open challenge.
Expert Commentary
The article presents a groundbreaking contribution to the field of turbulent flow simulation by introducing a functional, optimal transport-guided flow matching framework. The integration of optimal transport theory with functional data analysis in Hilbert space is particularly noteworthy, as it addresses a fundamental limitation of traditional generative models in handling infinite-dimensional data. The deterministic, straight-line probability paths not only enhance computational efficiency but also provide a theoretically elegant solution to the challenges of multi-scale intermittency in turbulent flows. The rigorous evaluation across diverse chaotic systems, including Navier-Stokes and Kolmogorov Flow, underscores the method's robustness and generality. However, the practical implementation of FOT-CFM may pose significant challenges due to the complexity of functional data handling and the computational demands of Hilbert space operations. Future work should focus on simplifying the framework for broader adoption and validating its performance in real-world, large-scale applications. This work sets a new benchmark for generative modeling in turbulence and signals a paradigm shift toward functional and geometric approaches in scientific computing.
Recommendations
- ✓ Develop open-source toolkits or libraries that abstract the complexities of Hilbert space operations, making FOT-CFM more accessible to practitioners in turbulence modeling and other scientific domains.
- ✓ Explore hybrid approaches that combine FOT-CFM with traditional grid-based methods to leverage the strengths of both paradigms, particularly in scenarios where functional data may be scarce or noisy.
- ✓ Conduct further validation studies in real-world applications, such as climate modeling or aerospace engineering, to assess the framework's scalability and robustness beyond benchmark systems.
Sources
Original: arXiv - cs.LG