Hyperparameter tuning in GANs

Hi all,

I am using GANs for generating synthetic MR images. I want to optimize some hyperparameters such as learning rate and batch size. I heard that performance of GANs are highly affected by hyperparameters. I think to optimize these hyperparameters using Bayesian Optimization in order to obtain best possible hyperparameter values in a relatively shorter time than using random or grid search. However, I couldn’t find an example using Bayesian Optimization for hyperparameter tuning for GANs. Do you think this is a bad idea? Which other methods do you use for hyperparameter optimization for GANs?

Thanks.