Problem
Research questions and friction points this paper is trying to address.
Wireless Physical Neural Networks
Nonlinearity
Linear Transformation
Expressiveness
Activation Function
Relay-Assisted Activation-Integrated SIM for Wireless Physical Neural Networks
🤖 AI Summary
Existing wireless physical neural networks are constrained by linear physical transformations, limiting their expressive power. This work proposes a relay-assisted multi-hop wireless physical neural network architecture that introduces trainable nonlinear processing capabilities in the analog domain by alternately cascading intelligent metasurfaces with hardware-implemented analog-domain nonlinear activation metasurfaces. The architecture jointly optimizes relay amplification matrices and programmable phase-shift matrices as network weights and, for the first time, integrates hardware-based activation functions directly into the physical layer. Simulation results demonstrate that the proposed approach substantially enhances the model’s nonlinear representational capacity and classification accuracy, significantly outperforming purely linear physical implementations.
📝 Abstract
Wireless physical neural networks (WPNNs) have emerged as a promising paradigm for performing neural computation directly in the physical layer of wireless systems, offering low latency and high energy efficiency. However, most existing WPNN implementations primarily rely on linear physical transformations, which fundamentally limits their expressiveness. In this work, we propose a relay-assisted WPNN architecture based on activation-integrated stacked intelligent metasurfaces (AI-SIMs), where each passive metasurface layer enabling linear wave manipulation is cascaded with an activation metasurface layer that realizes nonlinear processing in the analog domain. By deliberately structuring multi-hop wireless propagation, the relay amplification matrix and the metasurface phase-shift matrices jointly act as trainable network weights, while hardware-implemented activation functions provide essential nonlinearity. Simulation results demonstrate that the proposed architecture achieves high classification accuracy, and that incorporating hardware-based activation functions significantly improves representational capability and performance compared with purely linear physical implementations.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Wireless Physical Neural Networks
Activation-Integrated Metasurfaces
Nonlinear Analog Processing
Relay-Assisted Architecture
Trainable Physical Layer
🔎 Similar Papers
No similar papers found.
