How to use a customized Loss function in a Shallow neural network
24 views (last 30 days)
Hello everyone. There are very useful functions like trainlm and traingd to train a NN, but they all use MSE loss function as default. How can I change that with a function MyLossFunction that I build?
I searched a bit but I am finding basically solution to program all network, when the command train was very quicker. And even if I have to do step by step, I found how to build a network, do the forward propagation, create my Loss function, but I miss the backpropagation part.
Thank you for the attention
Srivardhan Gadila on 23 Aug 2020
In case of Shallow Neural Networks you can refer to Custom Performance Function for Neural Networks Toolbox using +mse package & relevant MATLAB Answers.
If deep neural networks would help for your use case then you can refer to Define Custom Training Loops, Loss Functions, and Networks & Train Network Using Custom Training Loop