How can I change the regularization (weight decay) in fitnet?

I use fitnet to create a 50-node model, with training and validation ratios set to 30% and 70%. I want the regularization (weight decay) set at 0.1 and 0.5 and and then find mean and variance of MSEs for training and validation portions of the dataset from the 10 repetitions. How can I do it?

Answer 1

[Binaya]

Hi Bao

To use regularization in "fitnet", you can modify the regularization performance parameter of the network. The following code snippet shows how to use regularization:

net = fitnet(50);
net.performParam.regularization = 0.1; % or 0.5

This parameter can be initialized alongside other parameters defined to configure the network. The rest of the provided code remains the same and the network can be trained to calculate the mean MSEs as per your requirement.

You can refer to the following links to understand about:

1. Properties of shallow neural networks: https://www.mathworks.com/help/deeplearning/ref/network.html?searchHighlight=shallow%20networks&s_tid=srchtitle_support_results_4_shallow%20networks#:~:text=given%20property%20values-,Properties,-Architecture%20Properties
2. Neural network object properties: https://www.mathworks.com/help/deeplearning/ug/neural-network-object-properties.html