Sequential Neural Probabilistic Amplitude Shaping: Learning the Channel's Language

2026-05-27

AI Summary

Traditional probabilistic amplitude shaping (PAS) methods often suffer from significant rate loss and low transmission efficiency when accounting for practical implementation penalties. This work proposes a block-free, sequence-level autoregressive neural encoder combined with arithmetic distribution matching to realize neural PAS. For the first time, the proposed scheme achieves a compelling balance between high performance and deployability under comprehensive consideration of all real-world implementation losses. By substantially reducing rate loss, it attains a higher achievable information rate than existing state-of-the-art approaches under realistic system constraints.

Abstract

We present the first neural probabilistic amplitude shaping that outperforms existing methods while accounting for all implementation losses, using a block-less, easily implementable sequential autoregressive encoder compatible with arithmetic distribution matching, yielding reduced rate loss and higher achievable information rates.

Problem

• probabilistic amplitude shaping
• neural sequential encoding
• information rates
• implementation losses
• arithmetic distribution matching

Innovation

• neural probabilistic amplitude shaping
• sequential autoregressive encoder
• arithmetic distribution matching
• rate loss reduction
• achievable information rates