Top-1 accuracy %: 75.53CRD training setting: ✘
Papers with Code · CC BY-SA 4.0
Knowledge Distillation
Knowledge distillation is the process of transferring knowledge from a large model to a smaller one. While large models (such as very deep neural networks or ensembles of many models) have higher knowledge capacity than small models, this capacity might not be fully utilized.
논문
Top-1 accuracy %: 77.03CRD training setting: ✘
: 70.70
Top-1 accuracy %: 70.81
Top-1 accuracy %: 71.61
Top-1 accuracy %: 71.96CRD training setting: ✓
Top-1 accuracy %: 71.38
Top-1 accuracy %: 71.63
Top-1 accuracy %: 71.72
Top-1 accuracy %: 71.61CRD training setting: ✓
Top-1 accuracy %: 70.66CRD training setting: ✓
Top-1 accuracy %: 76.376model size: 27MCRD training setting: ✘
Top-1 accuracy %: 72.19CRD training setting: ✘
Top-1 accuracy %: 72.39CRD training setting: ✘
Top-1 accuracy %: 71.68model size: 11.69MCRD training setting: ✓
Top-1 accuracy %: 72.07CRD training setting: ✘
Top-1 accuracy %: 71.62CRD training setting: ✘
Top-1 accuracy %: 72.04
Top-1 accuracy %: 71.38
Top-1 accuracy %: 71.87CRD training setting: ✓