ZENO: Nonlinear filtering with Zakai Equation Neural Operators
What it is
ZENO is a neural operator model that approximates the unnormalized filtering distribution of latent stochastic dynamical systems by optimizing against a physics-informed loss derived from the Zakai equation. It uses this inferred posterior within an expectation-maximization framework to perform state inference and parameter learning for stochastic differential equations.
The problem it solves
Solving high-dimensional stochastic partial differential equations (like the Zakai equation) for nonlinear filtering is analytically impossible and numerically intractable using traditional methods.
What you could use it for
- Inferring hidden states in nonlinear dynamical systems
- Learning parameters of stochastic differential equations from noisy observations
- Analyzing whole-brain imaging data from C. elegans to recover locomotion limit cycles
Why it matters
Provides a scalable, physics-informed method for solving high-dimensional nonlinear filtering problems that are analytically intractable, enabling accurate state estimation and parameter learning in complex scientific and engineering systems.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.