Graduate School and Research Center in Digital Sciences

Robust decentralized joint precoding using team deep neural network

de Kerret, Paul; Gesbert, David

ISWCS 2018, 15th International Symposium on Wireless Communication Systems, 28-31 August 2018, Lisbon, Portugal

Using Deep Neural Networks (DNNs) to tackle socalled Team Decision problems where the nodes aim at maximizing an expected common utility on the basis of different individual observations without any additional communications was recently introduced in a previous work and illustrated in the simple case of decentralized scheduling. In this work1, we apply this idea to design a decentralized robust precoding scheme in a Network MIMO configuration, which appears as a more challenging setting due to the continuous decision space and the required fine granularity of the precoding, in particular at high SNR. While the application remains fundamentally decentralized due to the decentralized nature of the channel state information (CSI), the training is done jointly. This is possible thanks to the common knowledge of the statistics (or equivalently the training data set) at all cooperating TXs. The joint training is done directly with respect to the desired figure-of-merit such that there is no need to generate labels using another method, and the precoding scheme obtained from the training does not only replicate a known scheme but is able to outperform state-of-the-art methods, as illustrated by simulations.

Document Doi Bibtex

Title:Robust decentralized joint precoding using team deep neural network
Type:Conference
Language:English
City:Lisbon
Country:PORTUGAL
Date:
Department:Communication systems
Eurecom ref:5600
Copyright: © 2018 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
Bibtex: @inproceedings{EURECOM+5600, doi = {http://dx.doi.org/10.1109/ISWCS.2018.8491209}, year = {2018}, title = {{R}obust decentralized joint precoding using team deep neural network}, author = {de {K}erret, {P}aul and {G}esbert, {D}avid}, booktitle = {{ISWCS} 2018, 15th {I}nternational {S}ymposium on {W}ireless {C}ommunication {S}ystems, 28-31 {A}ugust 2018, {L}isbon, {P}ortugal}, address = {{L}isbon, {PORTUGAL}}, month = {08}, url = {http://www.eurecom.fr/publication/5600} }
See also: