Gaussian flow dynamics improve learning of time-dependent stochastic models
Gaussian Flow Dynamics: Simulation-Free Neural SDE Learning Beyond One-Time Marginals
Machine Learning
Summary
Learning how systems change over time with randomness is tricky when you only know snapshots of their states at single moments. The authors show that just knowing these snapshots doesn't tell you the full story of how the system evolves. They introduce a new way to model these systems that explicitly accounts for the paths they take, not just the points they hit. This results in better learning of the underlying dynamics without needing to simulate many possible paths, saving time and computational effort.
What this means in practice
- •For machine learning engineers: Train models of time-dependent random processes with complex noise without costly simulations, improving efficiency and flexibility.
- •For financial modeling teams: Model asset price dynamics that have time-varying volatility using improved latent stochastic methods that capture more realistic temporal behaviors.
Authors
Grigory Bartosh, Christian A. Naesseth
Abstract
Simulation-free training of latent Stochastic Differential Equations (SDEs) relies on a variational posterior process whose one-time marginals are tractable, typically Gaussian. Such marginals, however, do not determine the underlying dynamics: many processes share the same marginals while differing in their temporal structure, and existing parameterizations fix this structure implicitly, which restricts the posterior family and biases the learned model. We introduce Gaussian flow dynamics, which construct stochastic processes directly from smoothly evolving Gaussian marginals while making the marginal-preserving, or gauge, degrees of freedom explicit and parameterizable. The construction admits state-dependent diffusion coefficients and recovers every linear SDE with additive noise and a non-degenerate Gaussian initial distribution. Building on it, we propose Gauge Matching, a simulation-free method for latent SDE learning that combines Gaussian flow dynamics with the SDE Matching objective. Gauge Matching costs at most quadratically in the latent dimension per step, like SDE Matching, but learns the temporal structure of the posterior beyond its one-time marginals. It comes within a nat of Helmholtz-SDE, which computes the gauge from the prior Jacobian at cubic cost, on the linear benchmark where the exact posterior is known, matches it on nonlinear systems, and applies where Helmholtz-SDE does not, to state-dependent noise.