trainNetwork - protect centain layers from changing

Hi,

I am trying to do train a neural net with my own layers but I dont want to change weights and biases in certain layers.

algorithm looks like this:

train Net1 with Layers1

train Net2 with Layers = [Layers2 Layers1]

My problem is that I dont want to anyhow alter weights and biases in Layers1, is there any way to do this?

Answer 1

[BERGHOUT Tarek]

yes there is a pssibility take this codes and change the training lope, when you find ''for i=1: nember of layes '' change it to for " i=2:nember of layes ", good luck.

https://www.mathworks.com/matlabcentral/fileexchange/69947-back-propagation-algorithm-for-training-an-mlp?s_tid=prof_contriblnk