Support Size of ε-Capacity-Achieving Inputs for the Amplitude-Constrained AWGN Channel

No prior work had resolved the exact scaling behavior of support sizes for capacity-achieving distributions under strict amplitude constraints. Researchers have long understood that optimal inputs consist of a finite set of discrete mass points. However, the precise growth rate of these points relative to increasing signal amplitude remains poorly defined. This gap motivated a shift toward analyzing inputs that achieve capacity within a small margin. Previous studies relied heavily on numerical simulations to estimate these scaling laws. Those observations often yielded conflicting results regarding the relationship between signal range and input complexity. That uncertainty drove the need for a more rigorous analytical framework. This paper addresses the theoretical limitations of existing models by focusing on relaxed optimality criteria. The investigation provides a structured way to interpret previously observed numerical data.

The aim of this work is to determine the minimal support size required for discrete inputs to achieve capacity within a small epsilon-gap. Researchers seek to resolve the scaling behavior of these inputs in the high-signal-to-noise ratio regime. This problem remains challenging because the exact capacity-achieving distribution is difficult to characterize analytically. The authors address the limitations of existing models by introducing a relaxed formulation of the optimality condition. They intend to provide a more tractable approach to understanding input complexity under amplitude constraints. By investigating various vanishing-gap regimes, the team hopes to clarify how signal range influences the number of mass points. This study is motivated by the need to reconcile conflicting scaling laws observed in previous numerical simulations. The investigation ultimately strives to offer a conceptual framework that explains the diversity of findings in the field.

The review approach involves a theoretical analysis of discrete-time channels under strict signal range limitations. Investigators utilize approximation theory to bound the performance of Gaussian mixture distributions. They apply information-theoretic techniques to manage entropy calculations through specific divergence metrics. A geometric wrapping procedure maps the channel input problem onto circular distribution approximations. This strategy facilitates the derivation of sharp characterizations for various vanishing-gap regimes. The team evaluates scaling behaviors as the signal amplitude constraint tends toward infinity. They contrast these analytical results with existing numerical data to validate their framework. This systematic methodology allows for the exploration of both polynomial and exponential gap decay scenarios.

The strongest finding establishes that for polynomially decaying gaps, the support size scales as theta of A log A as amplitude tends to infinity. This result holds for gaps defined by epsilon equals A to the power of negative beta with beta at least one. For exponentially small gaps, the authors derive bounds ranging between A log A and A to the power of three halves. The analysis confirms that the relaxed formulation is significantly more tractable than the exact capacity problem. These bounds provide a rigorous mathematical description of input complexity in the high-signal-to-noise ratio regime. The researchers show that their framework successfully captures the behavior of near-optimal distributions. Their results offer a unified explanation for the diverse scaling laws reported in previous computational literature. The findings demonstrate that input complexity is highly sensitive to the chosen approximation margin.

The authors demonstrate that relaxing the optimality requirement allows for precise characterizations of input complexity. Their analysis confirms that for polynomially decaying gaps, the required support size grows at a rate of A log A. This finding provides a theoretical basis for the scaling laws observed in earlier computational experiments. The researchers propose that varying regimes of approximation explain the diverse results seen in past literature. Their work suggests that exact optimizers may not be the sole drivers of observed numerical patterns. By linking the problem to circular distribution approximations, they offer a new perspective on information-theoretic control. The study clarifies why different numerical approaches might yield distinct scaling behaviors. These insights help unify disparate findings within the field of amplitude-constrained channel communication.