Clear Filters
Clear Filters

It is possible to implement a LSTM neural network built with Keras Python in a Simulink block?

6 views (last 30 days)
Hello!
I have a LSTM neural network (for time series prediction) built in Python with Keras. So I have the model (structure and weights) in .h5 file.
It is possible to implement a LSTM neural network built with Keras Python in a Simulink block?
I would to create a simulink file that takes in input 2 signals and passing through the NN block , gives me as output the predicted signal.
  2 Comments
Szabolcs Mike
Szabolcs Mike on 6 May 2019
Same question here too! I would like to use a LSTM prediction model trained with keras in Simulink. I can import the model to Matlab workspace, but not able to use it in a Matlab function block in Simulink!

Sign in to comment.

Answers (2)

Oskar Karnblad
Oskar Karnblad on 11 Jan 2018
I'm also interested in this. Have you found a solution yet?

Arkadiy Turevskiy
Arkadiy Turevskiy on 18 Oct 2021
You can do it in two steps:
  1. Import a model into MATLAB. See this function for TensrFlow 2 or this function to import from Keras.
  2. Once you have imported the model, use Stateful Predict block. This block became available in 21a.
HTH.
Arkadiy

Categories

Find more on Deep Learning with Simulink in Help Center and File Exchange

Community Treasure Hunt

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

Start Hunting!