How to use a trained LSTM network in MATLAB 2013a

I have an LSTM model trained on Matlab 2020b. I have a system where Matlab 2013a is installed. I cant upgrade the version. As of my understanding, 2013a does not support the deep learning toolbox, which is needed to use LSTM layers in neural networks. Is it possible to use the trained LSTM model in 2013a? Probably not directly, but what would be a workaround?

Thanks in advance

[Walter Roberson]

Can you do code generation on the R2020b system?

Answer 1

[Krishna]

Hi Tayyip,

If you wish to execute an LSTM model in MATLAB version 2013a, you can utilize code generation. The initial step involves converting your LSTM code written in MATLAB into C/C++ code. In MATLAB 2013a, you can employ the generated C/C++ code as an independent library by linking it with your MATLAB code. Subsequently, you can invoke the LSTM model functions from within your MATLAB 2013a code. This approach enables you to execute your model in the 2013a version. Please go through this documentation to learn more about code generation:

https://in.mathworks.com/help/coder/ref/codegen.html

Hope it helps,

Krishna