Federated Deep learning for CSI estimation in Massive MIMO environments

Abstract

Machine learning methods are poised to drastically improve the performance of many aspects of communication engineering, across all layers of the communication stack: from the physical layer to the application one. In this competition, we focus on the problem of federated training of a deep CSI compressor for massive MIMO in 5G protocols and beyond.

Instructions:

Objective:

A set of remote users observe a set of pilot signals as transmitted by a MIMO base station (BS) and are tasked with the distributed training of a compressor for the channel estimate. The training of this compressor occurs in a distributed manner, with the BS orchestrating the training and maintaining a centralized model. Training must occur within a set communication budget and model size.

Data:

The data is generated at https://ieee-dataport.org/open-access/ultra-dense-indoor-mamimo-csi-dataset

Competition specs:

https://docs.google.com/document/d/1m60WuDv2xcwhHeGtgwsG8iZFUA35z7UXlaaF...

Participants can discuss with the organizers on Slack here:
https://join.slack.com/t/itw2022federa-bzl1204/shared_invite/zt-1hutpzcx...