Let's Run Jinyeah

[논문] Distilling the Knowledge in Neural Network Knowledge Distillation의 초창기 논문 (2014 NIPS workshop) one model compression technique for bringing computations to edge devices. Where the goal is to have a small and compact model to mimic the performance of the cumbersome model Supervised learning (Label 존재) Response based knowledge (Teacher의 soft target) + Offline distillation (pre-trained Teacher..