How can you program Simulink to run multiple times, each of them using the output data from the previous run as initial condition?

I have implemented a radial basis neural network by hand in Simulink as a part of an adaptive approximation scheme. However, in order to train the network, I currently have to adjust the weights manually by changing an input array of weights located in the InitFcn callback. After every run, I update this array to the weights obtained in the run, which results in better predictions for the following run (i.e the network is trained). Given the time and effort that this requires, I am looking for a way to automate this process. Any suggestions would be greatly appreciated!

Answer 1

[Krishna]

Hi Nicolas,

I suggest utilizing the 'newrbe' or 'newrb' MATLAB commands to generate the Simulink model for radial basis neural networks instead of creating it manually. These commands offer greater flexibility in building the neural network and can enhance the training process, eliminating the need for repetitive manual training. For further information and detailed documentation, please refer to the following links:

https://in.mathworks.com/help/deeplearning/ref/newrbe.html

https://in.mathworks.com/help/deeplearning/ref/newrb.html

Additionally, you can use “genism” to convert it into a Simulink model.

https://in.mathworks.com/help/deeplearning/ref/network.gensim.html

Hope this helps.