The Capacity Region of the One-Sided Gaussian Interference Channel

Mojtaba Vaezi, H. Vincent Poor

TL;DR

This paper fully characterizes the capacity region of the one-sided Gaussian interference channel in the weak interference regime, a problem that has remained open for 40 years. The authors achieve this by developing a novel, tight outer bound and a new, simplified representation of the Han-Kobayashi inner bound, providing a complete understanding of achievable communication rates.

What Problem Does This Paper Address?

Imagine two people trying to communicate simultaneously, but one person's message is accidentally interfering with the other's. This is the essence of an interference channel in communication networks. Specifically, the "one-sided" version means only one of the two communication links experiences this interference. For decades, researchers have been trying to figure out the absolute maximum rate at which information can be reliably transmitted in such a scenario, especially when the interference is "weak" (i.e., the interfering signal is significantly weaker than the intended signal). While some progress has been made for "strong" interference or for the sum of the rates, the precise capacity region (the set of all possible pairs of reliable communication rates) for the weak interference case has remained elusive. This paper aims to close that gap.

Methods

The authors tackle this problem by employing a two-pronged approach:

  • A New Inner Bound (Achievable Region): They start by refining the well-known Han-Kobayashi (HK) inner bound. The HK scheme is powerful because it allows transmitters to split their messages into "private" (for their intended receiver only) and "common" (decodable by both) parts. The authors provide a simplified, yet powerful, representation of this HK bound for the one-sided Gaussian interference channel, explicitly identifying the optimal power split between private and common information.
  • A New Tight Outer Bound (Unachievable Region): To complement the inner bound, they develop a novel outer bound. This involves formulating a new optimization problem and proving that jointly Gaussian inputs are optimal for maximizing certain information-theoretic quantities. They then derive an extremal inequality that is crucial for establishing the tightness of this outer bound in the weak interference regime.

By combining these tight inner and outer bounds, they are able to precisely define the capacity region.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The "Weak" Interference Puzzle Solved: The paper finally provides a complete characterization of the capacity region for the one-sided Gaussian interference channel when interference is weak. This is a significant theoretical achievement.
  • Optimizing the Han-Kobayashi Scheme: The authors offer a clearer understanding of the Han-Kobayashi scheme by identifying the optimal power allocation strategy for private and common messages, making this powerful but complex technique more tractable.
  • A New Tool for Network Information Theory: The paper introduces a novel extremal inequality that proved instrumental in deriving the tight outer bound. This inequality is likely to be useful in solving other related problems in network information theory.
  • Unified View of Interference: The work also presents a unified theorem that covers both weak and strong interference regimes, offering a more complete picture of the one-sided interference channel's capacity.

Key Results

The paper establishes the exact capacity region for the one-sided Gaussian interference channel in the weak interference regime. This region is defined by a set of inequalities that precisely bound the achievable rates for both users. The authors demonstrate that their derived outer bound is tight, meaning it matches the achievable rates from their refined inner bound, thus completely characterizing the capacity. They also show that their results can be unified with existing results for the strong interference regime.

Limitations and Future Work

The authors acknowledge that while their work provides a complete characterization, the representation of the capacity region might still be complex. They suggest that the original Han-Kobayashi region might be preferred for its more intuitive presentation, even though their new formulation is more mathematically tractable. Future work could explore extending these results to more general interference channel models or to scenarios with more than two users. The newly developed extremal inequality also opens avenues for research in other network information theory problems.

Summary updated

Abstract

The capacity region of the one-sided Gaussian interference channel is established in the weak interference regime. To characterize this region, a new representation of the Han-Kobayashi inner bound for the one-sided Gaussian interference channel is first given. Next, a new outer bound on the capacity region of this channel is introduced which is tight in the weak interference regime. This is the first capacity region for any variant of the interference channel in the weak interference regime.
BibTeX
@misc{vaezi2015capacity,
  author = {Mojtaba Vaezi and H. Vincent Poor},
  title = {{The Capacity Region of the One-Sided Gaussian Interference Channel}},
  year = {2015},
  eprint = {1509.02200},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1509.02200},
}
Figure