Federated Deep learning for CSI estimation in Massive MIMO environments

Submission Dates:
10/14/2022 to 10/24/2022
Citation Author(s):
Sibren
De Bast
KU Leuven
Sofie
Pollin
KU Leuven
Submitted by:
Shih-Chun Lin
Last updated:
Mon, 10/24/2022 - 04:43
DOI:
10.21227/7png-jb17
License:
Creative Commons Attribution
1366 Views
Categories:

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...