Variational Structure at the Edge of Stability

• URL: http://arxiv.org/abs/2608.21660v1
• Date: Fri, 21 Aug 2026 22:05:12 GMT
• Title: Variational Structure at the Edge of Stability
• Abstract summary: We show that the edge coupling can be identified with the symmetric Verlet action.<n>We also show that the edge coupling can be identified with the symmetric Verlet action.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: When discrete-time optimizers operate at the edge of stability, they exhibit near-two-periodic behavior. These oscillatory dynamics are reminiscent of conservative systems, such as the dynamics generated by symplectic integrators. However, a precise formulation of the connection between discrete-time optimizers at the edge of stability and discrete mechanics remains underexplored. Recently, Litman introduced the "edge coupling": a functional on consecutive gradient descent iterates whose critical points encode the fixed points and two-point orbits of the gradient descent dynamics. Here we extend the edge coupling to heavy-ball and Nesterov momentum. We show that its critical points characterize the fixed points and two-point orbits, with its Hessian characterizing their stability. We also show that the edge coupling can be identified with the symmetric Verlet action, formalizing the connection between the edge of stability and discrete mechanics.

Related papers

• The Road Taken: The Role of Optimizers at the Edge of Stability [53.81268610971847]
  edge of stability refers to a phenomenon in deep learning where the Hessian eigenvalues of the loss remain stable above a threshold that the classical descent predicts to be unstable.<n>Previous works formulate the edge of stability with respect to the maximum Hessian eigenvalue and the learning rate.<n>We observe that many first-order methods, including gradient descent, significantly violate the stability bound predicted by these theories by a factor as large as $times 21.1$.
  arXiv  Detail & Related papers  (2026-08-19T01:03:34Z)
• Decoupling band topology from criticality in bosonic systems [0.0]
  A new understanding of criticality in systems described by quadratic bosonic Hamiltonians (QBHs) ties the emergence of long-range correlations to boundaries of dynamical stability.<n>We investigate the interplay of this landscape of possibilities with band topology and boundary physics.
  arXiv  Detail & Related papers  (2026-07-01T18:00:08Z)
• Momentum Further Constrains Sharpness at the Edge of Stochastic Stability [0.5405981353784005]
  Recent work suggests (stochastic) gradient descent self-organizes near an instability boundary, shaping both optimization and the solutions found.<n>We demonstrate that SGD with momentum exhibits an Edge of Stability (EoSS)-like regime with batch-size-dependent behavior.
  arXiv  Detail & Related papers  (2026-04-15T17:28:46Z)
• Stabilizing correlated pair tunneling of spin-orbit-coupled bosons in a non-Hermitian driven double well [4.706122486304247]
  We present an analytical framework for stabilizing second-order correlated tunneling of two bosons in a periodically driven non-Hermitian double-well potential.<n>Our analysis reveals distinct stability mechanisms for three fundamental channels: interwell spin-conserving, interwell spin-flipping, and intrawell spin-flipping.<n>These results expand the possibilities for controlling correlated tunneling in many-body systems with engineered dissipation.
  arXiv  Detail & Related papers  (2026-03-18T06:39:48Z)
• Beyond Scalars: Evaluating and Understanding LLM Reasoning via Geometric Progress and Stability [32.78289130870447]
  TRACED is a framework that assesses reasoning quality through theoretically grounded geometric kinematics.<n>We reveal a distinct topological divergence: correct reasoning manifests as high-progress, stable trajectories, whereas hallucinations are characterized by low-progress, unstable patterns.
  arXiv  Detail & Related papers  (2026-03-11T03:58:43Z)
• Convergence Rates for Gradient Descent on the Edge of Stability in Overparametrised Least Squares [33.60489399178793]
  gradient descent on neural networks is frequently performed in a large step size regime called the edge of stability"<n>We provide convergence rates for gradient descent with large learning rates in an overparametrised least squares setting.
  arXiv  Detail & Related papers  (2025-10-20T13:02:41Z)
• Observation of pattern stabilization in a driven superfluid [0.567478971466118]
  We observe the stabilization of square patterns in an interaction-driven, two-dimensional Bose-Einstein condensate.<n>These patterns emerge due to inherent high-order processes that become relevant in the regime of large phonon occupations.<n>We experimentally probe the predicted flows towards such a stable fixed-point, as well as repulsion from a saddle fixed-point.
  arXiv  Detail & Related papers  (2023-09-07T15:43:18Z)
• Beyond the Edge of Stability via Two-step Gradient Updates [49.03389279816152]
  Gradient Descent (GD) is a powerful workhorse of modern machine learning. GD's ability to find local minimisers is only guaranteed for losses with Lipschitz gradients. This work focuses on simple, yet representative, learning problems via analysis of two-step gradient updates.
  arXiv  Detail & Related papers  (2022-06-08T21:32:50Z)
• Clean two-dimensional Floquet time-crystal [68.8204255655161]
  We consider the two-dimensional quantum Ising model, in absence of disorder, subject to periodic imperfect global spin flips. We show by a combination of exact diagonalization and tensor-network methods that the system can sustain a spontaneously broken discrete time-translation symmetry. We observe a non-perturbative change in the decay rate of the order parameter, which is related to the long-lived stability of the magnetic domains in 2D.
  arXiv  Detail & Related papers  (2022-05-10T13:04:43Z)
• Feedback-induced instabilities and dynamics in the Jaynes-Cummings model [62.997667081978825]
  We investigate the coherence and steady-state properties of the Jaynes-Cummings model subjected to time-delayed coherent feedback. The introduced feedback qualitatively modifies the dynamical response and steady-state quantum properties of the system.
  arXiv  Detail & Related papers  (2020-06-20T10:07:01Z)
• Non-equilibrium non-Markovian steady-states in open quantum many-body systems: Persistent oscillations in Heisenberg quantum spin chains [68.8204255655161]
  We investigate the effect of a non-Markovian, structured reservoir on an open Heisenberg spin chain. We establish a coherent self-feedback mechanism as the reservoir couples frequency-dependent to the spin chain.
  arXiv  Detail & Related papers  (2020-06-05T09:16:28Z)
• Deconstructing effective non-Hermitian dynamics in quadratic bosonic Hamiltonians [0.0]
  We show that stability-to-instability transitions may be classified in terms of a suitably generalized $mathcalPmathcalT$ symmetry. We characterize the stability phase diagram of a bosonic analogue to the Kitaev-Majorana chain under a wide class of boundary conditions. Our analysis also reveals that boundary conditions that support Majorana zero modes in the fermionic Kitaev chain are precisely the same that support stability in the bosonic chain.
  arXiv  Detail & Related papers  (2020-03-06T19:30:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.