Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
To resolve inefficient straight-line endpoint coupling in Functional Flow Matching (FFM) without resorting to geometry-agnostic Euclidean metrics, this work develops kernel Functional Flow Matching (kFFM), which couples prior and target functions via entropic optimal transport under a Hilbert Sinkhorn Divergence kernel cost. The authors prove that the objective is uniformly bounded and well-posed on Banach ambient spaces, establishing a discretization-invariance bound whose convergence rate is governed by Sobolev regularity. Across time-series and turbulent Navier-Stokes benchmarks, kFFM consistently outperforms standard FFM, diffusion, and finite-dimensional $L^2$ Sinkhorn baselines, with specialized signature and Sobolev kernels yielding the largest gains on rough, path-valued data.
Function space flow matching stumbles when treating continuous fields like flat Euclidean vectors, but pairing endpoints via Hilbert Sinkhorn Divergence unlocks provably discretization-invariant generation across complex PDE systems.
Generative models for function-valued data, such as time series and solutions of partial differential equations, must learn distributions over infinite-dimensional spaces. Functional Flow Matching (FFM) extends Flow Matching to this setting, learning a velocity field whose flow transports a Gaussian prior to the data distribution, but it inherits the independent endpoint pairing of standard Flow Matching: in each batch, prior and data samples are matched arbitrarily, so the conditional bridge must traverse both the shared global structure of the dataset and instance-specific residuals. In function space this is harder to fix than in finite dimensions, since optimal transport (OT) on function spaces is delicate to formulate and a flat Euclidean surrogate ignores the geometry that distinguishes function-valued data. We propose kernel Functional Flow Matching (kFFM), which replaces the independent pairing by entropic OT under a kernel-induced cost, the coupling underlying the Hilbert Sinkhorn Divergence (HSD), leaving the FFM neural-operator architecture unchanged. We prove that the kernel cost and the HSD objective are uniformly bounded and well-posed on Banach ambient spaces, derive an error decomposition against quadratic-cost OT on compact metric spaces that isolates an irreducible kernel-cost mismatch term, and prove a discretization-invariance bound whose rate is governed by Sobolev regularity. Empirically, kFFM improves distributional matching over FFM, diffusion, adversarial, and finite-dimensional OT baselines on time-series and PDE benchmarks, with significant paired-seed gains over FFM and improvements that persist under non-kernel and physics-based diagnostics, including a turbulent Navier-Stokes benchmark. Bounded kernel costs already outperform raw $L^2$ Sinkhorn, and function-space-aware kernels (signature, Sobolev RBF) give further gains on rough or path-valued data.