SupplementaryMaterial: ProgressiveKernelBasedKnowledgeDistillationfor AdderNeuralNetworks

In this section, more experimental results of PKKD are conducted. Then, we show the superiority of the proposed methods on the traditional CNN distillation. Model Top-1acc Top-5acc ResNet-18 69.8% 89.1% PKKD 73.1% 91.3% VanillaKD[1] 72.5% 90.9% Finally, we show the experimental results of using different settings of PKKD on ImageNet with ResNet-50inTab.3.