you can using nntool

Thank you for your answer.

I already checked the nntool from the Neural Network Toolbox, but I was not able to found how to train the NN using a custom functional that I defined.

Apparently, using this tool you can only choose from several predefined training algorithms, such as Levenberg-Marquardt, Resilient Backpropagation or Conjugate Gradient. In these algorithms the loss function to update the weights of the NN relies on the error between the network output and the target, thus it are not suited for my particular problem (I do not have any target to train with).

Could you explain me how can I define a custom functional (not a transfer function) to update the weighs of the NN in nntool?

you can use in weka tool

Ok, thank you for your answer.

Hi

You should use S-funtion.

Thank you for your answer.

As far as I know, S-funtion is used inside the Simulink environment. However, I'm working outside simulink, by means of .m files. Even so, do you know how can I train a NN with a custom loss funtion inside an S-Function?

The mathworks weblink give u sufficient information:

1. https://in.mathworks.com/help/deeplearning/ref/train.html

2. https://in.mathworks.com/help/deeplearning/ug/choose-a-multilayer-neural-network-training-function.html

3. https://in.mathworks.com/help/deeplearning/ug/train-and-apply-multilayer-neural-networks.html

4. https://in.mathworks.com/help/deeplearning/ug/try-deep-learning-in-10-lines-of-matlab-code.html

Useful video lecture on Neural Network toolbox

https://www.youtube.com/watch?v=2Z4959acjKs