Co-execute MATLAB and Keras to simulate the effect of channel estimate compression on precoding in a MIMO OFDM channel.
Updated 10 Feb 2023
Simulate Channel Compression Effect on Precoding Using MATLAB and Keras with CSINet
This example shows how to co-execute MATLAB® and Python® to simulate the effect of channel estimate compression on precoding in a MIMO OFDM channel. It shows how to:
  • Generate CDL channel estimates in MATLAB using the 5G Toolbox™
  • Load and test a pre-trained CSINet Keras™ model using co-execution with Python
  • Fine tune the model weights through transfer learning using co-execution with Python
  • Simulate the effect of channel estimate compression on precoding in MATLAB using the 5G Toolbox and Communications Toolbox™
The following figure summarizes the operations executed in MATLAB (marked in blue) versus those executed in Keras (marked in grey).
There are different options for accessing deep learning models within MATLAB, including:
  1. Using models created in MATLAB using Deep Learning Toolbox™
  2. Converting models from other frameworks into MATLAB
  3. Co-executing models from other frameworks with MATLAB
This example provides an overview of the third approach. Co-execution is useful for leveraging wireless products in MATLAB to test existing deep learning models from other frameworks in an end-to-end link simulation. This workflow also allows engineering teams working with MATLAB & Python deep learning frameworks to easily combine their work in one environment. Approaches one & two are useful for building deep learning models from scratch or for using additional capabilities from the Deep Learning Toolbox with pre-trained models, such as Deep Learning Code Generation or deep learning models in Simulink®.
To run this example, you need:
For more information about installing Python, see Install Supported Python Implementation.
Running the Example
Open and run the live script SimChanCompEffOnPrecodingUsingMATLABAndKerasWithCSINetExample.mlx. To generate a new CDL channel estimates dataset, use the live script GenerateCSINetDataSet.mlx. The live scripts use the helper files preprocessChannelEstimate.m and postprocessChannelEstimate.m for pre/post-processing the training and testing data sets.Visualize the channel compression effect
By the end of the example, you will be able to visualize the effect of channel feedback compression with CSINet [1] on the received constellation of an OFDM MIMO channel with zero-forcing precoding. The following figure shows the effect of using CSINet for channel feedback with compression rate 1/4 and normalized mean square error -42 dB in a CDL channel with the following parameters:
  • Tx Antennas: 32
  • Rx Antennas: 2
  • Delay Profile: CDL-B
  • RMS delay spread: 100 ns
  • Max delay after truncation: 32
  • Max Doppler: 2 Hz
  • Resource blocks: 48
  • Subcarrier spacing: 30 KHz
[1] Wen, Chao-Kai, Wan-Ting Shih, and Shi Jin, "Deep learning for massive MIMO CSI feedback," IEEE Wireless Communications Letters, vol. 7, no. 5, pp. 748-751, Oct. 2018.
Copyright 2023, The MathWorks, Inc.

Cite As

Maha Fadel (2024). CSINet-Channel-Compression-in-MATLAB-Using-Keras (https://github.com/matlab-deep-learning/CSINet-Channel-Compression-in-MATLAB-Using-Keras/releases/tag/v1.0.0), GitHub. Retrieved .

MATLAB Release Compatibility
Created with R2022b
Compatible with R2022b and later releases
Platform Compatibility
Windows macOS Linux
Tags Add Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!
Version Published Release Notes

See release notes for this release on GitHub: https://github.com/matlab-deep-learning/CSINet-Channel-Compression-in-MATLAB-Using-Keras/releases/tag/v1.0.0

To view or report issues in this GitHub add-on, visit the GitHub Repository.
To view or report issues in this GitHub add-on, visit the GitHub Repository.