Learning to Write on Dirty Paper

2025-07-23

AI Summary

This work addresses the practical challenge of dirty paper coding (DPC) where the transmitter has perfect channel state information but lacks prior knowledge of input distribution, interference statistics, and channel statistics. We propose an end-to-end learnable neural encoder–decoder framework that bypasses explicit modeling assumptions about channel or interference distributions. By jointly optimizing nonlinear neural encoders and decoders via supervised learning, the framework automatically learns effective nonlinear precoding mappings. Notably, it is the first purely data-driven approach to recover core structural features of Tomlinson–Harashima precoding and lattice coding—without imposing analytical constraints. Experimental results demonstrate substantial performance gains over classical DPC schemes across diverse channel and interference scenarios, achieving robust interference pre-cancellation without requiring any statistical priors.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Dirty paper coding (DPC) is a classical problem in information theory that considers communication in the presence of channel state known only at the transmitter. While the theoretical impact of DPC has been substantial, practical realizations of DPC, such as Tomlinson-Harashima precoding (THP) or lattice-based schemes, often rely on specific modeling assumptions about the input, state and channel. In this work, we explore whether modern learning-based approaches can offer a complementary path forward by revisiting the DPC problem. We propose a data-driven solution in which both the encoder and decoder are parameterized by neural networks. Our proposed model operates without prior knowledge of the state (also referred to as "interference"), channel or input statistics, and recovers nonlinear mappings that yield effective interference pre-cancellation. To the best of our knowledge, this is the first interpretable proof-of-concept demonstrating that learning-based DPC schemes can recover characteristic features of well-established solutions, such as THP and lattice-based precoding, and outperform them in several regimes.

Problem

• Explores learning-based solutions for dirty paper coding
• Proposes neural network encoder-decoder without prior knowledge
• Demonstrates learning outperforms traditional DPC methods

Innovation

• Neural networks parameterize encoder and decoder
• Operates without prior channel or state knowledge
• Recovers nonlinear mappings for interference pre-cancellation