Forward-Forward Autoencoder Architectures for Energy-Efficient Wireless Communications
Daniel Seifert, Onur Günlü, Rafael F. Schaefer
TL;DR
This paper addresses the energy and memory limitations of backpropagation in learning-based wireless communication systems. It introduces forward-forward autoencoders trained with contrastive inputs, enabling end-to-end learning without differentiable channels and with layer-wise goodness metrics. The results show FF autoencoders can closely match or surpass BP-based performance on AWGN and Rayleigh block fading channels, particularly when a non-differentiable quantization stage is enforced, and they offer memory and processing-time savings due to the removal of the backward pass. These findings indicate FF learning as a viable route toward energy-efficient neural codes suitable for practical wireless hardware and non-differentiable channels.
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.