Pre-trained models and datasets for "Fairness-Aware Low-Rank BD-RIS Design via Curriculum Self-Supervised Learning"

This dataset contains channel realizations and pre-trained neural network checkpoints used to reproduce the figures of the paper "Fairness-Aware Low-Rank BD-RIS Design via Curriculum Self-Supervised Learning" submitted to IEEE Wireless Communications Letters. Contents:- PaperWCL_sim.mat: 1000 channel realizations (3GPP TR 38.901 UMa, 2.6 GHz) used for evaluating Figures 4 and 5.- paperWCL.mat: 100,000 channel realizations used to generate the training dataset.- TrainingDataRepo.rar: archive containing the PyTorch dataset (Dataset_M_10_N_30_K_5_Ntx_2_realiz_100000_self_supervised.pt) and the pre-trained checkpoints (full-rank, low-rank, no-curriculum, and fairness-aware variants with λ ∈ {0.1, 0.3, 0.5, 0.7}). System configuration: M=10 BS antennas, N=30 BD-RIS elements, K=5 users, Ntx=2 transmit antennas per user. Companion code: https://github.com/DarielPereira/Low-Rank-BDRIS