ff1418e8cc993fe8abcfe3ce2003e5c5-AuthorFeedback.pdf

Forthe training cost, itindeed depends on the ratio ofthe passport-aware branch activated inevery training8 epoch. The default ratio value is 50%, i.e., train 1 iteration passport-aware branch after training every 1 iteration9 passport-free branch. In this case, the theoretical computation cost will be 2x. In SE, the transformation is to learn a channel-wise attention17 to enhance the feature representation. We adopt a similar setting as the trigger-set based method [10].