1 paper · 1 filter
Logan Frank, Jim Davis
Knowledge distillation (KD) has been a popular and effective method for model compression. One important assumption of KD is that the original training dataset is always available.…