Multiplicative Capacity Gains from Feedback in Multiuser Channels

35 concepts25 papers2 proposals

In classic information theory, a foundational result established by Claude Shannon is that perfect feedback does not increase the capacity of a memoryless point-to-point channel. While feedback can simplify code design, it cannot push the fundamental data rate beyond the channel's physical limits.

However, this intuition completely breaks down in multi-user networks. When multiple transmitters and receivers share a medium, feedback ceases to be a redundant echo and instead becomes a powerful catalyst for cooperation, unlocking multiplicative and potentially unbounded capacity gains.

How Feedback Neutralizes Interference

To understand why feedback is so revolutionary in multi-user settings, consider the two-user Gaussian interference channel, where Transmitter 1's signal interferes with Receiver 2, and vice versa. Without feedback, each pair must treat the other's signal as destructive noise or carefully coordinate their independent codebooks to avoid collisions.

When feedback is introduced, a new mechanism emerges. As shown by Changho Suh et al. in Feedback Capacity of the Gaussian Interference Channel to within 2 Bits, feedback allows a receiver to send its received signal back to the opposing transmitter. This transmitter can then deduce the exact interference its competitor is experiencing. Armed with this knowledge, the transmitters can act as virtual relays for each other's messages, actively aligning or canceling interference at the receivers.

The consequence of this cooperation is profound: the capacity benefit scales proportionally with signal power. In high-interference regimes, this multiplicative gain becomes unbounded as the signal-to-noise ratio (SNR) increases, as further characterized by Alireza Vahid et al. in Interference Channels with Rate-Limited Feedback. This stands in stark contrast to point-to-point channels, where feedback offers only negligible or constant additive improvements.

This capacity-boosting effect is not limited to isolated two-user pairs. In On the Feedback Capacity of the Fully Connected $K$-User Interference Channel, Soheil Mohajer et al. demonstrate that feedback enables a "cooperative interference alignment" strategy that scales efficiently even as the number of interfering users grows. Furthermore, Naveen Goela demonstrates in Approximate Feedback Capacity of the Gaussian Multicast Channel that these unbounded gains are a fundamental property of complex, many-to-many network connectivity rather than a quirk of specific interference topologies.

Reshaping the Degrees of Freedom

The structural impact of feedback is most clearly visible in the Generalized Degrees of Freedom (GDoF) region, which measures how capacity scales in high-power regimes. Without feedback, the GDoF of a symmetric interference channel plotted against interference strength forms a notorious "W"-shaped curve. The dip in the middle of the "W" represents a frustrating regime where interference is too strong to be ignored but too weak to be efficiently decoded and stripped away.

Feedback fundamentally alters this landscape. In On the Capacity Region and the Generalized Degrees of Freedom Region for the MIMO Interference Channel With Feedback, Mehdi Ashraphijuo et al. demonstrate that for symmetric Multiple-Input Multiple-Output (MIMO) interference channels, the presence of feedback transforms the GDoF region into a "V"-shaped curve. By enabling a completely different trade-off in interference management, feedback eliminates the inefficient middle regime of the "W" curve, allowing capacity to be characterized within a constant number of bits regardless of signal strengths.

GDoF
^
|  \     /      \     /
|   \   /        \   / 
|    \_/          \_/
|     No Feedback (W-shape)
+--------------------------> Interference Strength

GDoF
^
|  \             /
|   \           /
|    \         /
|     \_______/
|     With Feedback (V-shape)
+--------------------------> Interference Strength

This structural expansion also manifests in decentralized environments. Samir Medina Perlaza et al. show in Perfect Output Feedback in the Two-User Decentralized Interference Channel that feedback strictly expands the set of stable operating points ($\eta$-Nash equilibrium regions), allowing selfish users to achieve rate pairs much closer to the social optimum.

Overthrowing the Zero-Sum Assumption

A common skepticism regarding feedback in practical systems is that it must be a "zero-sum game": any bandwidth or power allocated to sending feedback signals must surely detract from the system's forward data rate.

However, interactive communication models prove otherwise. In Two-way interference channels, I-Hsiang Wang et al. model the feedback path not as a free pipe, but as a realistic, resource-constrained channel. They show that allocating a portion of the channel for feedback yields a net increase in total system capacity because the data rate gained in the forward direction significantly outweighs the bandwidth sacrificed in the backward direction.

Expanding on this, Changho Suh et al. demonstrate in Two-way interference channel capacity: How to have the cake and eat it too that in certain regimes, two interfering links can achieve the maximum possible "perfect feedback" capacity in both directions simultaneously—essentially gaining the benefits of feedback for free without sacrificing transmission rates.

Furthermore, this feedback does not need to be perfect or continuous to be effective:

  • Intermittent Feedback: Can Karakus et al. show in Gaussian Interference Channel with Intermittent Feedback that even highly unreliable feedback provides multiplicative GDoF gains. Transmitters can treat the intermittent feedback as a quantization problem ("quantize and forward"), achieving near-perfect feedback scaling laws even when the feedback link is active only half the time.
  • Periodic Feedback: In time-varying environments, Mehdi Ansari Sadrabadi et al. demonstrate in On The Capacity Of Time-Varying Channels With Periodic Feedback that periodic feedback of channel state information allows optimal adaptation using a single Gaussian codebook by dynamically adjusting power.

Beyond Data Transmission: Infinite Identification Capacity

The most extreme manifestation of feedback's power occurs when the communication goal shifts from traditional message transmission (decoding a specific bitstream) to message identification (determining whether a specific message of interest was sent).

In Identification via Gaussian Multiple Access Channels in the Presence of Feedback, Yaning Zhao et al. explore Gaussian Multiple Access Channels (GMAC) with noiseless feedback. They discover that feedback enables an infinite identification capacity. Under this paradigm, the number of identifiable messages can grow faster than any double-exponential function of the code length. Feedback allows the transmitters to dynamically exploit the shared channel noise, transforming what is normally a destructive force into a coordinating mechanism that completely bypasses traditional Shannon capacity limits.

Context-Dependent Gains

While feedback is immensely powerful, its utility is highly context-dependent and tied to the underlying channel physics:

  • Channel Degradation: Venugopal Veeravalli notes in Cooperative relay broadcast channels that feedback does not increase capacity for degraded channels, but significantly improves it for non-degraded channels by enabling better user cooperation.
  • Function Computation: In On Modulo-Sum Computation over an Erasure Multiple Access Channel, Ashish J Khisti et al. show that causal feedback of the channel state significantly boosts the capacity for computing functions (like modulo-sums) over erasure multiple access channels, a benefit that does not carry over to standard point-to-point erasure links.
  • State Uncertainty: Feedback can resolve internal state uncertainties. Dor Shaviv et al. show in Can Feedback Increase the Capacity of the Energy Harvesting Channel? that feedback strictly increases capacity when a transmitter relies on stochastic energy harvesting, allowing it to manage its finite battery based on what the receiver has actually observed.
  • Network Routing: In full-duplex cellular networks, Kaiming Shen et al. demonstrate in Capacity Limits of Full-Duplex Cellular Network that feedback mechanisms enabling a downlink user to receive information about uplink transmissions via a base station significantly expand the achievable rate region.

Ultimately, across all these diverse multi-user settings, feedback transforms receivers from passive destinations into active participants, turning shared interference and noise into resources for network-wide coordination.

Key Foundational Work

Cooperative relay broadcast channels (2005) — contributed 2 concept elements to this direction.
On The Capacity Of Time-Varying Channels With Periodic Feedback (2007) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_unbounded_feedback_gain_93bff284,
  title = {unbounded feedback gain},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/unbounded-feedback-gain/rdr_93bff28468e3473f90e9c736464fed5a},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_unbounded_feedback_gain_93bff284_v1,
  title = {unbounded feedback gain},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/unbounded-feedback-gain/rdr_93bff28468e3473f90e9c736464fed5a/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake