This paper studies the capacity of massive random-access networks modeled as a multiple-input multiple-output fading channel with infinitely many interfering users. The network is assumed to operate in a noncoherent regime, where transmitters and receivers know the fading statistics but not their realizations. Users access the network via random activation with a given probability. To characterize the symmetric sum rate, a random-coding argument is invoked together with the assumption that users and interferers draw their codebooks according to the same distribution. For this channel model, rigorous upper and lower bounds on the network capacity are derived. The behavior of these bounds depends critically on the spatial decay of the large-scale fading statistics from interfering users. In particular, if the large-scale fading coefficients of the interferers (ordered according to their distance to the receiver) decay exponentially or more slowly, then the capacity is bounded in the transmit power. This occurs because the aggregate interference scales with the transmit power, and reveals an inherent saturation effect in interference-limited networks. Moreover, in this regime, random user activity cannot fundamentally eliminate the resulting capacity ceiling. In contrast, if the large-scale fading coefficients of the interferers decay faster than double-exponentially, then the capacity becomes unbounded in the transmit power. Note that proving an unbounded capacity is nontrivial even if the number of interfering users is finite, since the condition that the users’ codebooks follow the same distribution prevents interference-avoiding strategies such as time-, frequency-, or code-division multiple access, and cooperation among users associated with different access nodes is not allowed. An unbounded coding rate is achieved by using bursty signaling together with treating interference as noise.

This paper proposes a comprehensive framework for optimizing the deployment of Unmanned Aerial Vehicles (UAVs) in a communication system, aiming to minimize the transmission power of UAVs while ensuring the required data rates for ground users. The proposed model accounts for both large- and small-scale fading effects. To accurately capture the characteristics of the wireless channel, the Rician Shadowed fading model is adopted, as it incorporates both line-of-sight (LoS) and nonline-of-sight (NLoS) components. A mathematical formulation of the problem is presented, and the impact of the Rician shadowed fading parameters, the severity of the fluctuations in the LoS component, and the ratio between the powers of the LoS and NLoS components is considered. The simulation results demonstrate that increasing the parameters improves channel conditions, thus reducing the transmission power required to meet the minimum data rate constraints. To address the UAV 3D placement problem, which results in a non-convex and analytically intractable optimization due to the nonlinearities introduced by the channel model. Therefore, we employ a heuristic algorithm, Particle Swarm Optimization Evolutionary (PSO-E).

Despite advances in deep probabilistic models, learning discrete latent representations remains challenging. This work introduces a novel method to improve inference in discrete Variational Autoencoders by reframing the inference problem through a generative perspective. We conceptualize the model as a communication system, and propose to leverage Error-Correcting Codes (ECCs) to introduce redundancy in latent representations, allowing the variational posterior to produce more accurate estimates and reduce the variational gap. We present a proof-of-concept using a Discrete Variational Autoencoder with binary latent variables and low-complexity repetition codes, extending it to a hierarchical structure for disentangling global and local data features. Our approach significantly improves generation quality, data reconstruction, and uncertainty calibration, outperforming the uncoded models even when trained with tighter bounds such as the Importance Weighted Autoencoder objective. We also outline the properties that ECCs should possess to be effectively utilized for improved discrete variational inference.

The channel capacity of wireless networks is often studied under the assumption that the communicating nodes have perfect channel-state information (CSI) in the sense that they have access to the fading coefficients in the network. To the best of our knowledge, one of the few works that studies wireless networks without this assumption is by Lozano, Heath, and Andrews. Inter alia, Lozano et al. show that in the absence of perfect CSI, and if the channel inputs are given by the square-root of the transmit power times a power-independent random variable, then the achievable information rate is bounded in the signal-to-noise ratio (SNR). However, such inputs do not necessarily achieve capacity, so one may argue that the information rate is bounded in the SNR because of the suboptimal input distribution. In this paper, it is demonstrated that if the nodes do not cooperate and they all use the same codebook, then the achievable information rate remains bounded in the SNR even if the input distribution is allowed to change arbitrarily with the transmit power.