The Rate-Distortion-Perception-Classification Tradeoff: Joint Source Coding and Modulation via Inverse-Domain GANs
This addresses the problem of efficient image compression and transmission for communication systems, presenting a novel tradeoff and algorithms that improve performance over existing methods, though it is incremental in advancing deep learning-based joint coding frameworks.
The paper tackles the joint source coding and modulation problem by revealing a strict tradeoff between channel rate, distortion, perception, and classification accuracy (RDPC), and proposes two algorithms, including an inverse-domain GAN method, that achieve extreme compression and outperform traditional and recent deep methods in metrics like distortion, perception, and classification accuracy.
The joint source-channel coding (JSCC) framework leverages deep learning to learn from data the best codes for source and channel coding. When the output signal, rather than being binary, is directly mapped onto the IQ domain (complex-valued), we call the resulting framework joint source coding and modulation (JSCM). We consider a JSCM scenario and show the existence of a strict tradeoff between channel rate, distortion, perception, and classification accuracy, a tradeoff that we name RDPC. We then propose two image compression methods to navigate that tradeoff: the RDPCO algorithm which, under simple assumptions, directly solves the optimization problem characterizing the tradeoff, and an algorithm based on an inverse-domain generative adversarial network (ID-GAN), which is more general and achieves extreme compression. Simulation results corroborate the theoretical findings, showing that both algorithms exhibit the RDPC tradeoff. They also demonstrate that the proposed ID-GAN algorithm effectively balances image distortion, perception, and classification accuracy, and significantly outperforms traditional separation-based methods and recent deep JSCM architectures in terms of one or more of these metrics.