Hi Mohammad,
In MATLAB, ‘fitnet’ and ‘network’ are both functions that can be used to create and train neural networks, but they differ in their flexibility and ease of use.
- ‘fitnet’ : It is a high level function that provides a convenient way to create and train a neural network with a single function call.
- ‘network’: This can be used to create customized shallow neural networks, it is a low level function that provides more control over the network architecture, and we can set the ‘trainingOptions’ in detail, so more work is required to setup and train network compared to ‘fitnet’
Please, refer to the following documentations for further details,
Please, refer to the following documentation, to understand how to edit the Shallow neural network properties:
https://www.mathworks.com/help/deeplearning/ug/create-and-train-custom-neural-network-architectures.html
As for the learning rate, you can specify it using ‘trainingOptions’ object, refer to the following example,
Example:
% Create a shallow neural network with one hidden layer
net = feedforwardnet([10]);
% Set the training options
options = trainingOptions('trainlm', 'MaxEpochs', 50, 'LearningRate',0.01);
% Train the neural network using the specified options
net = train(net, inputs, targets, options);
And, it is not possible to use mini-batch learning with the ‘traingd’ Optimizer. The 'traingd' optimizer is a batch gradient descent algorithm that updates the network weights using the gradients computed over the entire training dataset.But,it is possible to use mini-batch learning with other optimizers when working with shallow neural networks in MATLAB. The ‘train’ function supports mini-batch learning with the 'trainscg' and 'traingdm' optimizers, which are both based on gradient descent.
Hope this helps!
