Channel-Estimation-ML-DOJO-2

This code present our results to the ML5G-PHY [channel estimation] challenge, which is part of the ITU Artificial Intelligence/Machine Learning in 5G Challenge of 2020.

Problem statement:

The problem statement can be found here: Challenge: Site-specific channel estimation with hybrid MIMO architectures
The train and evaluation data can also be downloaded from that link.

Report:

Our report presenting the development of the algorithms and training procedures is available here.

Code:

Requirements:

The code has been developed using Matlab 2019b.

Description:

• TD1SNR1_example_for_test_data.m is an example of how to obtain the predictions for test data for the test dataset 1- SNR1.

• ReconstructChannel.m is the function with our algorithm to obtain the channels.
  Whitening.m is a subfunction used to whitten the noise on ReconstructChannel.m.

• Train.m is the script used for training to obtain the detection threshold.
  ReconstructChannel_train.m is the function similar to ReconstructChannel.m used during the training script.

Results:

The test channels can be downloaded from this drive folder