Does Matlab classification learner app allow changes to hyperparameters such as epoch and learning rate?

Hi,

i currently exploring both machine learning and deep learning in Matlab. I notice that when i try to train CNN in deep learning, i could modify the epoch, learning rate and batch size in trainingOptions such as code below.

options = trainingOptions('sgdm', ...
    'MiniBatchSize',miniBatchSize, ...
    'MaxEpochs',3, ...
    'InitialLearnRate',1e-4, 
    'Plots','training-progress');

However, when i using classification learner app the UX is better but there is no options for me to tune the hyperpareters such as epoch as the training is done automatically and very fast.

So back to my questions, is there any setting that allow me to tune the hyperparameters such as epoch and learning rate in classification learner app?

Hope someone could help to clarify. Thank you.

Answer 1

[Abhijeet]

Hi, Teo

I understand that you wanted to explore tuning hyperparameters in the Classification learner application.

Classification learner application provides an option called "optimizer" though which you can tune the hyper-parameters.

You can find the option "optimizer" in the top panel.

To understand the various parameters available for tuning, you can kindly refer to below mentioned documentation:

https://in.mathworks.com/help/stats/hyperparameter-optimization-in-classification-learner-app.html?searchHighlight=learning%20rate%20classification%20learner%20app&s_tid=srchtitle_learning%20rate%20classification%20learner%20app_3