Forward-Forward Autoencoder Architectures for Energy-Efficient Wireless Communications

2025-10-13

AI Summary

Traditional end-to-end learned communication autoencoders based on backpropagation (BP) suffer from high energy consumption, large memory and computational overhead, and reliance on global gradients—particularly problematic in non-differentiable channels such as Rayleigh fading. To address these limitations, this paper proposes a lightweight autoencoder architecture leveraging Forward-Forward (FF) learning. The architecture eliminates the need for differentiable modulation schemes or explicit channel modeling, thereby circumventing BP’s dependence on global partial derivatives; it is specifically designed for joint coding and modulation, supporting discrete symbol mapping and non-differentiable channel simulation. Experimental results demonstrate that the FF-based autoencoder achieves bit-error-rate performance comparable to BP baselines under both AWGN and Rayleigh fading channels, while reducing training time by approximately 40% and memory footprint by 35%. This work constitutes the first systematic application of FF learning to wireless communication autoencoders, establishing a new paradigm for energy-efficient, edge-deployable intelligent transceivers.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Learning to Search
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

The application of deep learning to the area of communications systems has been a growing field of interest in recent years. Forward-forward (FF) learning is an efficient alternative to the backpropagation (BP) algorithm, which is the typically used training procedure for neural networks. Among its several advantages, FF learning does not require the communication channel to be differentiable and does not rely on the global availability of partial derivatives, allowing for an energy-efficient implementation. In this work, we design end-to-end learned autoencoders using the FF algorithm and numerically evaluate their performance for the additive white Gaussian noise and Rayleigh block fading channels. We demonstrate their competitiveness with BP-trained systems in the case of joint coding and modulation, and in a scenario where a fixed, non-differentiable modulation stage is applied. Moreover, we provide further insights into the design principles of the FF network, its training convergence behavior, and significant memory and processing time savings compared to BP-based approaches.

Problem

• Designing energy-efficient autoencoders using forward-forward learning for wireless communications
• Replacing backpropagation with forward-forward algorithm for non-differentiable channels
• Evaluating performance on Gaussian noise and Rayleigh fading communication channels

Innovation

• Forward-forward learning replaces backpropagation for training
• Autoencoders designed for non-differentiable communication channels
• Energy-efficient implementation with memory and time savings