Learning Dynamics in Memristor-Based Equilibrium Propagation
This addresses energy-efficient in-memory computing for AI hardware, but it is incremental as it builds on existing equilibrium propagation and memristor models.
The paper tackled the problem of how nonlinear memristor-based weight updates affect convergence in neural networks trained with equilibrium propagation, finding that robust convergence is achievable if memristors have a resistance range of at least an order of magnitude.
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.