The experiments in the paper should be based on CIFAR10 and Convnet, and the lr_img and lr_net are set for these two by default, right? If I want to switch to vgg11BN and ResNet18BN_AP, what should I adjust the learning rates to for better performance? Thank you.
The experiments in the paper should be based on CIFAR10 and Convnet, and the lr_img and lr_net are set for these two by default, right? If I want to switch to vgg11BN and ResNet18BN_AP, what should I adjust the learning rates to for better performance? Thank you.