Code of "Recursive CSI Quantization of Time-Correlated MIMO Channels by Deep Learning Classification", IEEE SPL 2020
Contact: Stefan Schwarz, Institute of Telecommunications, TU Wien, stefan.schwarz@tuwien.ac.at
This code can be used to reproduce the neural network quantization results of
"Recursive CSI Quantization of Time-Correlated MIMO Channels by Deep Learning Classification", S. Schwarz, IEEE SPL, 2020
The code is setup for a small-scale MIMO system with 4 transmit and 2 receive antennas, in order to speed-up the execution. However, these parameters can be changed in "Quant_example.m".
The code requires Matlab's Deep Learning Toolbox.
To run the code, execute the main file "Quant_example.m".
This file will call the scripts "NN_training.m" and "train_net.me" for DNN training.
Afterwards, "time_corr.m" will be executed to apply the trained multi-stage quantizer for quantization of time-correlated channels.