Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Abbott, Cliff B., Elo, Mark, Bozhko, Dmytro A.
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2506.04523
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915327407816704
author Abbott, Cliff B.
Elo, Mark
Bozhko, Dmytro A.
author_facet Abbott, Cliff B.
Elo, Mark
Bozhko, Dmytro A.
contents We introduce Perturbative Gradient Training (PGT), a novel training paradigm that overcomes a critical limitation of physical reservoir computing: the inability to perform backpropagation due to the black-box nature of physical reservoirs. Drawing inspiration from perturbation theory in physics, PGT uses random perturbations in the network's parameter space to approximate gradient updates using only forward passes. We demonstrate the feasibility of this approach on both simulated neural network architectures, including a dense network and a transformer model with a reservoir layer, and on experimental hardware using a magnonic auto-oscillation ring as the physical reservoir. Our results show that PGT can achieve performance comparable to that of standard backpropagation methods in cases where backpropagation is impractical or impossible. PGT represents a promising step toward integrating physical reservoirs into deeper neural network architectures and achieving significant energy efficiency gains in AI training.
format Preprint
id arxiv_https___arxiv_org_abs_2506_04523
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Perturbative Gradient Training: A novel training paradigm for bridging the gap between deep neural networks and physical reservoir computing
Abbott, Cliff B.
Elo, Mark
Bozhko, Dmytro A.
Machine Learning
Mesoscale and Nanoscale Physics
Emerging Technologies
Neural and Evolutionary Computing
Computational Physics
We introduce Perturbative Gradient Training (PGT), a novel training paradigm that overcomes a critical limitation of physical reservoir computing: the inability to perform backpropagation due to the black-box nature of physical reservoirs. Drawing inspiration from perturbation theory in physics, PGT uses random perturbations in the network's parameter space to approximate gradient updates using only forward passes. We demonstrate the feasibility of this approach on both simulated neural network architectures, including a dense network and a transformer model with a reservoir layer, and on experimental hardware using a magnonic auto-oscillation ring as the physical reservoir. Our results show that PGT can achieve performance comparable to that of standard backpropagation methods in cases where backpropagation is impractical or impossible. PGT represents a promising step toward integrating physical reservoirs into deeper neural network architectures and achieving significant energy efficiency gains in AI training.
title Perturbative Gradient Training: A novel training paradigm for bridging the gap between deep neural networks and physical reservoir computing
topic Machine Learning
Mesoscale and Nanoscale Physics
Emerging Technologies
Neural and Evolutionary Computing
Computational Physics
url https://arxiv.org/abs/2506.04523