Loss of Monotonicity in the Vector Play Model and Consequences for Newton-Type Solvers
Abstract
The vector play model (VPM) replaces the implicit minimization of the energy-based hysteresis model (EBM) by an explicit update. Its tangent contains the product of the Hessians of two unrelated potentials and is unsymmetric. Hence, the VPM has no co-energy density at frozen memory, and the skew part makes H↦B lose monotonicity in a band just outside the pinning circle, into which small time steps drive the sliding states. In composite models, the determinant of the tangent can change sign, which explains published failures of the local inversion B↦H, and where it is negative, H↦B is not injective. In the scalar potential formulation, the exact tangent keeps a descent guarantee, whereas symmetric surrogates of the tangent, such as the symmetrized tangent or Broyden-Fletcher-Goldfarb-Shanno (BFGS) updates, lose it and become ascent directions where the skew part dominates, as computations of TEAM Problem 32 confirm. It is therefore recommended to use solvers with an unsymmetric tangent.
Related
- Beyond PINNs: A Unified Gauss--Newton and Petrov--Galerkin Framework for Neural and Hybrid PDE Solvers
- Comparative Performance Analysis of Optimization Solvers for Linear Support Vector Machines: Impact of Preconditioning and Dimensionality An Empirical Study Using the Benchopt Framework
- Vector and matrix monotonicity characterisation of the "load forces" in nonlinear elasticity: compressible versus incompressible response
- Training-Loss Guarantees for Muon with Finite-Step Newton--Schulz Orthogonalization
- Explicit counterexamples to Frank-Loss optimizer conjectures for critical Dirac and vector Sobolev inequalities
BibTeX
@misc{cibicki2026loss,
title = {{Loss of Monotonicity in the Vector Play Model and Consequences for Newton-Type Solvers}},
author = {Dominik Cibicki and Klaus Roppert and Manfred Kaltenbacher},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.00168},
eprint = {2611.00168},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.00168},
}
RIS
TY - UNPB
TI - Loss of Monotonicity in the Vector Play Model and Consequences for Newton-Type Solvers
AU - Dominik Cibicki
AU - Klaus Roppert
AU - Manfred Kaltenbacher
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.00168
UR - https://arxiv.org/abs/2611.00168
AB - The vector play model (VPM) replaces the implicit minimization of the energy-based hysteresis model (EBM) by an explicit update. Its tangent contains the product of the Hessians of two unrelated potentials and is unsymmetric. Hence, the VPM has no co-energy density at frozen memory, and the skew part makes H↦B lose monotonicity in a band just outside the pinning circle, into which small time steps drive the sliding states. In composite models, the determinant of the tangent can change sign, which explains published failures of the local inversion B↦H, and where it is negative, H↦B is not injective. In the scalar potential formulation, the exact tangent keeps a descent guarantee, whereas symmetric surrogates of the tangent, such as the symmetrized tangent or Broyden-Fletcher-Goldfarb-Shanno (BFGS) updates, lose it and become ascent directions where the skew part dominates, as computations of TEAM Problem 32 confirm. It is therefore recommended to use solvers with an unsymmetric tangent.
ER -