Federated learning (FL) often struggles with generalization due to heterogeneous client data. Local models are prone to overfitting their local data distributions, and even transferable features can be distorted during aggregation. To address these challenges, we propose FedCONST, an approach that adaptively modulates update magnitudes based on the global model’s parameter strength. This prevents over-emphasizing well-learned parameters while reinforcing underdeveloped ones. Specifically, FedCONST employs linear convex constraints to ensure training stability and preserve locally learned generalization capabilities during aggregation. A Gradient Signal-to-Noise Ratio (GSNR) analysis further validates FedCONST's effectiveness in enhancing feature transferability and robustness. As a result, FedCONST effectively aligns local and global objectives, mitigating overfitting and promoting stronger generalization across diverse FL environments, achieving state-of-the-art performance.

System level performance of a broadcast/unicast superposition coding system is considered. Cellular network for unicast service only is considered as interference limited system, where increasing transmission power does not help improving network throughput. It is the case especially when frequency reuse factor is close to 1. In such case, the amount of power that does not contribute to improving throughput can be considered as "unused." This situation motivates to use the unused power for broadcast service, which can be efficiently provided in OFDM based single frequency network (SFN) as of the digital video/audio broadcast system. In this paper, we investigate the performance of such an overlay system where single frequency broadcast network is superimposed over unicast cellular service, namely, BC/UC superposition coded system (supercast). Alternative service multiplexing using FDM/TDM will also be investigated for comparison.

The problem of optimal power allocation for a single user in an OFDMA system is considered for uplink, under the assumption that the total power is divided equally to a selected subset of subcarriers. This scenario suggests reduced feedback information for power control from base station to the user since only the information over modulation on each subcarrier is necessary. Assuming knowledge of the instantaneous users' channel gains, the rate optimization problem reduces to a subcarrier allocation problem, described by a discrete function with the set of possible allocated subcarriers as domain and the set of achievable sum rates as range. A continuous equivalent of the function is used to derive properties that also hold in the discrete case. The function is in general not concave and concavity holds only for high CNR regions - over a specified bound. In all cases however it is proved that the maximum is unique and a simple and efficient algorithm is proposed for subcarrier allocation which always provides the optimal solution concerning the subcarriers to be loaded. The cost in achievable capacity from the use of equal-power compared to actual waterfilling is calculated, which is shown to be small while the gain in frequency resources and feedback reduction is important.

In this paper, strategies for dynamic user grouping and shared frequency resource assignments for OFDMA systems are proposed. The users are grouped based on their modulation type and a metric for this purpose is introduced which considers tradeoffs between maximum rate and minimum allocated resources per user. The available resources are distributed through the modulation groups based on fairness criteria. Equal power allocation on the loaded subcarriers of each user is further incorporated. Simulations show that under equal power assumptions the algorithm proposed in the following can bring a better exploitation of the available bandwidth and can provide good QoS to an increased number of users simultaneously.

In this paper, for spatial multiplexing with limited feedback, we propose quantized principal component selection (QPCS) precoding, wherein the active bases are selected at the receiver from a finite number of basis sets known at the both receiving and transmitting ends, conveyed to the transmitter using limited feedback, and assembled into a precoding matrix at the transmitter. The selected bases are conveyed to the transmitter using feedback information on both the index of the selected basis set, which defines the most appropriate set of coordinates for describing a multiple-input multiple-output (MIMO) channel, and the principal bases with the largest gain in the selected basis set. A systematic design method for a finite number of basis sets using the antenna array processing is presented. We show that the proposed QPCS precoding scheme provides comparable capacity with the closed-loop MIMO with full channel state information (CSI), even with limited feedback. Furthermore, for practical implementation of the QPCS precoding scheme, we propose a low complex structured precoder optimization procedure and show that the proposed procedure induces a negligible capacity loss compared with the exhaustive precoder optimization, even with considerably reduced complexity.

This paper studies the problem of efficient exploitation of the available bandwidth on the reverse link (RL) of cdma2000 1/spl times/EV-DV system. The objective is to maximize user throughput and to minimize packet delay while ensuring that forward link (FL) overhead is as low as possible. Two radio resource management (RRM) mechanisms, the time-scheduled (TS) scheme and the rate-controlled (RC) scheme, which can be used as the RRM schemes for the RL of 1/spl times/EV-DV are introduced and compared. It is shown that the RC scheme offers a better sector throughput through higher rise-over-thermal (RoT) utilization with less FL overhead, while the TS scheme outperforms the RC scheme in terms of packet delay through the capability of fast rate adaptation. In addition, a hybrid RRM scheme, namely rate control with quick start (RCQS) is introduced to combine the advantages of the TS and RC schemes. It is shown that the throughput and delay of the RCQS are comparable to those of the RC scheme the TS scheme, respectively.

The paper evaluates the performance of hybrid automatic repeat request (HARQ) for the code division multiple access (CDMA) reverse link in a power controlled scenario and assesses the effect of imperfect power control on HARQ. Analytical results on the performance of HARQ and the effect of imperfect power control are derived and verified using simulation in Rayleigh and Rician fading channels. Results show that HARQ provides robustness against imperfect power control.

This paper evaluates the performance of hybrid autonomous request (HARQ) for the cdma2000 reverse link in a power controlled scenario and assesses the effect of imperfect power control on HARQ. Analytical results on the performance of HARQ and the effect of imperfect power control are derived and verified using simulation in Rayleigh fading channels. Results show that HARQ provides robustness against imperfect power control.