AWhat changed
Combined continuous-time invertible dynamics with Hutchinson trace estimation and adjoint-based training to scale log-density evaluation without triangular neural Jacobian restrictions.
Scientific scope
Combined continuous-time invertible dynamics with Hutchinson trace estimation and adjoint-based training to scale log-density evaluation without triangular neural Jacobian restrictions. Inherits continuous normalizing flows, neural ODEs, adjoints and Hutchinson estimation. Unbiasedness concerns the stochastic log-density integral under stated regularity and exact integration; exponentiation is not thereby an unbiased density estimate and practical ODE solvers introduce numerical error. One-pass sampling is an ODE solve with potentially many function evaluations, not a single network evaluation.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits continuous normalizing flows, neural ODEs, adjoints and Hutchinson estimation. Unbiasedness concerns the stochastic log-density integral under stated regularity and exact integration; exponentiation is not thereby an unbiased density estimate and practical ODE solvers introduce numerical error. One-pass sampling is an ODE solve with potentially many function evaluations, not a single network evaluation.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction