CS Paper Radar

Explicit Second-Order Bounds on the Domain of Validity for Lyapunov-Schmidt Reduction

Pranav Gupta, Ravi Banavar, Anastasia Bizyaeva

eess.SY ·submitted 2026-09-21 14:56 UTC · v1
theoryunusual

Abstract

Lyapunov-Schmidt reduction is a widely used dimensionality reduction technique for bifurcation analysis in high-dimensional systems. While classical formulations guarantee the local existence of reduced-order equations, they typically lack explicit quantitative estimates on the size of the neighbourhoods in which these representations faithfully capture the full bifurcation structure. In recent work, we have addressed this limitation by deriving bounds on the domain of validity of this reduction using first-order conditions on the vector field together with a quantitative result for the implicit function theorem. In this article we explore beyond, and adopt second-order conditions that incorporate the Hessians of the vector field to develop these bounds, and obtain a new set of results. We then investigate the applicability of these newly derived bounds to Hopfield-like networked dynamical systems, evaluate these bounds for pitchfork bifurcations on connected regular graphs, and finally examine the relationship between the two certified domains for this class of systems.

Comment
This manuscript has been submitted for review to Elsevier's Systems & Control Letters

Closest papers

Nearest neighbours by meaning across the whole index.

ETM-3D: A One-Parameter Multistable Quadratic Flow with an Explicit Absorbing Set and Intermittent Chaos

eess.SY nlin.CDmath.DSmath.NT+1 ·2026-10-03 ·Roya Khalili-Amirabadi, Mohsen Jalaeian-Farimani
benchmarkunusual similarity 0.50
Abstract

We study a one-parameter three-dimensional quadratic flow, denoted ETM-3D, on the benchmark domain $ρ\geq0$. Its homogeneous quadratic triad preserves Euclidean quadratic energy, while the full vector field has constant divergence -8. A shifted quadratic Lyapunov function yields an explicit positively invariant absorbing set and therefore global ultimate boundedness. The equilibria reduce generically to a scalar quartic, with a separate singular case at $ρ=\sqrt{5}$. Continuation reveals two saddle-node bifurcations, a generic transcritical bifurcation, a supercritical Hopf bifurcation, and a fold of limit cycles. Stable and unstable period-one branches meet at the fold with their nontrivial Floquet multipliers approaching +1. Just beyond this point, a 25-point logarithmic parameter study gives a laminar-length scaling exponent of -0.551 (95% confidence interval [-0.621,-0.481]), consistent with type-I intermittency. At larger $ρ$, a chaotic attractor coexists with an asymptotically stable equilibrium and periodic windows occur. Structural invariants are used to distinguish the flow from several classical quadratic chaotic families.

0.26

Learning Morse Lyapunov functions through invertible neural networks

eess.SY math.OC ·2026-11-16 ·Wouter Jongeneel
similarity 0.49
Abstract · 13 pages, 3 figures, comments are very welcome! Currently under review

Towards data-driven stability certificates of dynamical systems, we are witnessing renewed interest in studying and exploiting the structure of Lyapunov functions. Especially Lyapunov functions that are locally quadratic (i.e., Morse) gained popularity ever since it was understood that they can be written as the norm squared of a diffeomorphism. This topological representation alludes to parametrizations using invertible neural networks. In this note we discuss when and how such functions can be learned. In particular, using neural ODEs (NODEs), we characterize Sobolev approximation of these diffeomorphisms on compact sets and show how this naturally relates to stability. We also connect our results to PLNets and supply numerical experiments for illustration.

0.31

A (Purely) Graph-Theoretic Approach to Synchronization of Nonlinear Dynamical Networks

eess.SY math.DSnlin.CD ·2026-08-18 ·Aandrew Baggio Sahaya Arokiadoss, G. Arunkumar
published similarity 0.48
Abstract · 10 page, 3 figures

Synchronizing nonlinear dynamical networks typically requires solving matrix inequalities or detailed system models, which fail for large networks. This paper offers a simple fix : a purely graph-theoretic framework using only a single Lipschitz-like bound on the dynamics. Coupling strengths are computed directly from the digraph, bypassing inequality solvers entirely. The method succeeds where existing approaches encounter infeasibility due to connectivity patterns. It examines only $n-1$ directed paths per strongly connected component versus $\frac{n(n-1)}{2}$ undirected paths before, achieving $O(n^3)$ complexity. Results show network connectivity can be exploited to synchronize a large class of nonlinear dynamical networks.

0.35

On global Lyapunov functions being Morse

eess.SY math.DSmath.OC ·2026-10-04 ·Wouter Jongeneel
unusual similarity 0.47
Abstract · 15 pages, 2 figures

Lyapunov functions have been a cornerstone of dynamical systems theory ever since the early 1900s. Moreover and akin to Morse theory, Lyapunov functions have been the key in linking topology to dynamical systems theory. In this note we continue along these lines. Suppose that some vector field on Euclidean space has a unique equilibrium point that is globally asymptotically stable (GAS). Under these conditions, it is known that there is always a smooth Lyapunov function to certify stability. One may wonder if this function can always be chosen to be Morse. We show that if the equilibrium point is hyperbolic, this is indeed true. Moreover, we show that if this would be true in general, then the generalized 4-dimensional smooth Poincaré conjecture (SPC4) must be true. Even so, we provide an example that is GAS, but does not admit a Morse Lyapunov function, closing this route of proving SPC4.

0.50

Petrov-Galerkin operator inference with application to stability-encouraging identification

eess.SY math.OC ·2026-10-01 ·Johannes Rettberg, Jonas Nicodemus, Harsh Sharma, Boris Kramer, Jörg Fehr +1 more
benchmark similarity 0.46
Abstract

Data-driven model order reduction methods such as operator inference enable the efficient construction of reduced-order models directly from high-dimensional time-domain data. Standard operator inference typically seeks a Galerkin-type reduced model in a prescribed low-dimensional subspace by identifying its reduced operators from projected snapshot data. The resulting inference problem is formulated as a least-squares problem admitting an efficient closed-form solution. However, it is well known from intrusive model order reduction for linear time-invariant systems that Petrov-Galerkin projections can additionally preserve important system properties such as stability and passivity. To overcome the limitations of standard operator inference, we extend the framework for linear time-invariant systems to incorporate Petrov-Galerkin projections and provide explicit error expressions and bounds between the intrusive and nonintrusive reduced operators, thus generalizing results from the literature. We demonstrate the proposed approach in the context of dissipative and port-Hamiltonian systems. Furthermore, we introduce a novel convex optimization formulation that explicitly enforces the port-Hamiltonian structure on the inferred operators. The effectiveness of the proposed methods is demonstrated on several well-established benchmark problems, including the CD player, an atmospheric model, a mass-spring-damper system, and a poroelasticity system.

0.45

Convex Equilibrium Potentials for Monge--Ampère: Curvature Bounds and A Posteriori Evaluation

cs.LG math.NA ·2026-10-22 ·Edison Vázquez, Yingdong Lu, Tomasz Nowicki
similarity 0.46
Abstract

Equilibrium neural networks define their hidden state as the fixed point of a contraction, so their effective depth is set at evaluation time rather than by the architecture. Using them for fully nonlinear PDEs raises two issues that explicit networks avoid: second derivatives may grow with the effective depth, and the potential is only ever evaluated approximately. We study nonnegative softplus equilibrium networks with contraction factor $q<1$, a convex class suited to the Monge--Ampère equation. A positivity cancellation in the resolvent shows that the Hessian grows at most like $(1-q)^{-2}$, rather than the $(1-q)^{-3}$ of a direct estimate, and an explicit one-neuron family shows that this exponent is sharp; third derivatives satisfy a sharp $(1-q)^{-3}$ bound. Residual-based a posteriori estimates then control the state, its input derivatives and, in two dimensions, the Hessian determinant at an approximate equilibrium. Combined with interval arithmetic on complete covers of the domain, they yield rigorous uniform bounds on the Monge--Ampère residual and, through a comparison argument, on the solution error of a fixed trained network. On two manufactured Dirichlet problems on the unit disk, all six trained models are certified with residual below $0.25$ and boundary error below $0.01$. Adaptive stopping halves float64 evaluation time at best, whereas certified adaptive evaluation remains costlier than fixed depth.

0.33

Stability Transfer for Autonomous Systems via Higher-Order Taylor-Truncated Lyapunov Certificates

eess.SY math.OC ·2026-10-13 ·Jose D. Hoyos, Zehui Lu, Shaoshuai Mou
published similarity 0.46
Abstract

We develop analytical sufficient conditions for certifying the local asymptotic stability of nonlinear autonomous systems based on Lyapunov functions derived from their finite-order Taylor polynomial approximations. This approach translates the stability analysis problem from the original nonlinear dynamics to a more computationally tractable polynomial domain. The derived conditions allow for the explicit calculation of a guaranteed inner approximation of the region of attraction ($\mathrm{ROA}$). We show how sum-of-squares (SOS) programming and grid search can be used to systematically search for polynomial Lyapunov functions that yield a large certified threshold radius and extract a certified invariant set inner approximation of the $\mathrm{ROA}$ of the nonpolynomial original system. The framework is then applied to a control tuning problem: finding reduced-effort Linear Quadratic Regulator (LQR) controllers that ensure that a given initial state of interest belongs to the certified $\mathrm{ROA}$ of the nonpolynomial closed loop system. The effectiveness of the overall methodology is demonstrated using a pendulum with nonlinear damping and a nonlinear control system.

0.29

Port-Hamiltonian Neural Networks for Systems with Multiple Asymptotically Stable Equilibria

cs.LG eess.SY ·2026-10-01 ·Simon Heilig, Jens Püttschneider, Mohammad Itani, Asja Fischer, Timm Faulwasser
publishedtheory similarity 0.45
Abstract · Accepted at NeurIPS 2026 Workshop: AXIOM - Foundations of Efficient Deep Learning

Stable port-Hamiltonian neural networks certify asymptotic stability by construction. Yet, their Hamiltonian is a global Lyapunov function with a single global minimum, so they can represent only dynamic systems with one attractor. We demonstrate that this excludes even simple systems with energy landscapes forming a double well, and we overcome the restriction by parametrising the Hamiltonian as a product of Bregman divergences generated by one input-convex network. We prove that the resulting model is locally Lyapunov stable, that the coexistence of stable equilibria forces additional non-asymptotically-stable equilibria to exist, that all equilibria lie in a bounded region, and under a hyperbolicity assumption that almost-everywhere stability holds. On three systems our approach is able to recover the energy surface characteristics and improve the convergence speed by 1.8$\times$-8.5$\times$.

0.26