Blow-up Parameter Landscapes for Polynomial Dynamical Systems

arXiv:2607.14269 2026 Dynamics 2 ideas extracted · analyzed Aug 30, 2026

What the math gives to ML

The paper develops a constructive method for detecting finite-time blow-up in parameterized polynomial dynamical systems by compactifying infinity and analyzing the resulting boundary flow. Its transferable asset is the reduction of large-state behavior to the highest-degree homogeneous component: directional equilibria at infinity satisfy F_n(u) = alpha u, and the sign of alpha together with angular stability predicts runaway trajectories. This can be used in polynomial neural ODEs, recurrent models, state-space models, and deep equilibrium systems as an instability atlas and as a runtime burst controller. The main falsifiable prediction is that hidden-state divergence begins near parameter values where an attracting direction at infinity acquires positive radial growth.

Ideas from this paper

Failed on benchmark 2026

Infinity Atlas for Polynomial Neural Flows

For a neural ODE, residual flow, or deep equilibrium model with a dominant polynomial component, compute the directional dynamics induced by its highest-degree homogeneous term on the unit sphere. Penalize or reject parameter regions containing radially growing attracting directions, preventing finite-time activation blow-up while preserving nonlinear dynamics in safe directions.

Useful8/10
Difficulty6/10
Novelty8/10
Paper: Blow-up Parameter Landscapes for Polynomial Dynamical Systems arXiv:2607.14269
Mechanism confirmed, baseline not beaten 2026

Compactified Burst Controller

Use the paper's distinction between radial attraction and tangential instability at infinity to detect impending hidden-state bursts before they cause numerical failure. When the state approaches a radially growing directional equilibrium, temporarily add radial damping or switch to a bounded fallback update, then restore the original dynamics after angular ejection.

Useful7/10
Difficulty5/10
Novelty7/10
Paper: Blow-up Parameter Landscapes for Polynomial Dynamical Systems arXiv:2607.14269