Artifacts for "Federated Learning for CSI Feedback in mMIMO: A Communication-Performance Trade-Off Analysis"

This repository contains the artifacts associated with the paper"Federated Learning for CSI Feedback in mMIMO: A Communication-Performance Trade-Off Analysis" (Duque, Hendrikx, Gorce - 2026). Contents - federated_csinet_dataset.mat: QuaDRiGa-generated dataset for 20 clients in Urban Macrocell (UMa) scenario (mixed indoor/outdoor, LOS/NLOS).- federated_csinet_dataset_Umi.mat: QuaDRiGa-generated dataset in Urban Microcell (UMi) scenario, used for pretraining.- pretrained_UMi.pt: CRNet model pretrained on the UMi dataset, used as warm-start initialization for federated fine-tuning experiments. Code - Model training: https://gitlab.inria.fr/lduque/declearncsi- Dataset generation: https://gitlab.inria.fr/lduque/fedcsi_quadriga-dataset