How to build a Custom neural network Layer (PG-JANET)

I was assigned to implement an algoritm from a paper ("Digital Predistortion of RF Power Amplifiers With Decomposed Vector Rotation-Based Recurrent Neural Networks") in Matlab. I'm familiar with neural network, mainly using Matlab toolbox (existing) layers. However, PG-JANET is not in the existing list.

I'll appreciate any help with this: similar examples and how to start the implementation.

Thank you,

Alecs

The layer is defined by:

And:

Answer 1

[Garmit Pant]

Hello Alecsander

I understand that you are trying to implement the algorithm given in the paper "Digital Predistortion of RF Power Amplifiers With Decomposed Vector Rotation-Based Recurrent Neural Networks" in MATLAB but the LSTM layer, PG-JANET, is not available in the list of the existing layers.

As of MATLAB 2023B, you can define custom LSTM layer in MATLAB. You need to define the layer as an M-File.

Please follow the following MATLAB example to define a custom LSTM layer.

https://www.mathworks.com/help/deeplearning/ug/define-custom-recurrent-deep-learning-layer.html

For further understanding, you can refer to the following MATLAB Documentation:

1. https://www.mathworks.com/help/deeplearning/ug/define-custom-deep-learning-intermediate-layers.html - Tutorial to set up a general deep learning custom layer.
2. https://www.mathworks.com/help/deeplearning/ug/long-short-term-memory-networks.html - Example to depict how to set up and use an LSTM neural network.

I hope this helps!

Best Regards

Garmit

[Alecsander Eitan]

Thank you for your answer.

This is a general answer, with links to some general examples. It is quite difficult to use these examples and tutorials for building a new and different layer.