how to choose the optimumize values of the training option

Dear all I have a question, I am using Transfer Learning of the pretraining Deep learning networks I would like to know how to choose the optimum values of the training options (Solver, initial learning rate, max epoch, miniBatchsize etc) should I keep trying different values or choosing the default values, please

Answer 1

[Udit06]

Hi,

You can consider using Bayesian optimization in order to find optimal network parameters. Bayesian optimization is an algorithm well suited to optimizing hyperparameters of classification and regression models.

You can refer to the following MathWorks documentation for more details related to Bayesian optimization:

https://www.mathworks.com/help/releases/R2021a/deeplearning/ug/deep-learning-using-bayesian-optimization.html

If Bayesian optimization does not suits your use case, you can consider other optimization techniques like grid search or randomized search optimization. You can refer to the following MathWorks documentation for more details:

https://www.mathworks.com/help/releases/R2021a/stats/hyperparameter-optimization-in-classification-learner-app.html#:~:text=Description-,Optimizer,-The%20optimizer%20values

I hope this helps.