mute others

please

Good afternoon, Professor Jin. During the training phase, e.g., CNN, we can quantitatively compare different models with their layer-wise gradient value or loss descent graphs. However, I am wondering how to mathematically evaluate their generalization, except those evaluation metrics on the testing data.

Good afternoon, Professor Jin. Thanks for sharing! In the medical imaging scene, sometimes it is difficult for us to obtain the ground truth. So, we use the simulated data to train the network. So, how can we better generalize the network trained on simulated data to real data?

Thank you