English

Learning Dynamics in Memristor-Based Equilibrium Propagation

Machine Learning 2026-02-18 v1 Emerging Technologies Neural and Evolutionary Computing

Abstract

Memristor-based in-memory computing has emerged as a promising paradigm to overcome the constraints of the von Neumann bottleneck and the memory wall by enabling fully parallelisable and energy-efficient vector-matrix multiplications. We investigate the effect of nonlinear, memristor-driven weight updates on the convergence behaviour of neural networks trained with equilibrium propagation (EqProp). Six memristor models were characterised by their voltage-current hysteresis and integrated into the EBANA framework for evaluation on two benchmark classification tasks. EqProp can achieve robust convergence under nonlinear weight updates, provided that memristors exhibit a sufficiently wide resistance range of at least an order of magnitude.

Keywords

Cite

@article{arxiv.2512.12428,
  title  = {Learning Dynamics in Memristor-Based Equilibrium Propagation},
  author = {Michael Döll and Andreas Müller and Bernd Ulmann},
  journal= {arXiv preprint arXiv:2512.12428},
  year   = {2026}
}
R2 v1 2026-07-01T08:23:36.916Z