Contour integral methods and structured perturbations for linear differential-algebraic equations
arXiv:2607.12628
2026
Dynamics
2 ideas extracted · analyzed Aug 30, 2026
What the math gives to ML
The paper offers a transferable method for evolving stiff or algebraically constrained neural state-space models without relying on stability-limited explicit time steps. Its central construction is inverse-Laplace propagation of descriptor systems, evaluated through a small set of complex resolvent solves. This can become a contour-integral SSM layer with reusable factorizations, or a constraint-preserving neural dynamical block in which algebraic relations hold throughout propagation rather than only through a penalty loss. The most promising validation targets are larger stable time steps, reduced constraint drift, and improved long-horizon gradient behavior.
Ideas from this paper
△ Mechanism confirmed, baseline not beaten
2026
Build a neural dynamical block whose hidden state contains differential variables and Lagrange multipliers, with a singular descriptor matrix enforcing constraints during propagation. This avoids the drift and ill-conditioning that can arise when exact constraints are represented only by a penalty term.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Failed on benchmark
2026
Replace repeated time-stepping of a stiff linear state-space block with a quadrature approximation to its inverse Laplace transform. The layer propagates a hidden state using a small set of complex shifted linear solves, which can be batched and reused across many time steps or parameter values.
Useful7/10
Difficulty6/10
Novelty6/10