概念
distillation training
创建2026-04-15
更新2026-04-15
阅读量级1 分钟
概念导读

Knowledge Distillation (KD), introduced by Hinton et al. in 2015, is the foundational technique for transferring knowledge from a large "teacher" model to a smaller "student" model. The student is trained to match the teacher's output distribution (typically via kl-divergence-in-distillation), preserving much of the teacher's capability at lower computationa...

  1. Related
  2. References

Knowledge Distillation

Knowledge Distillation (KD), introduced by Hinton et al. in 2015, is the foundational technique for transferring knowledge from a large "teacher" model to a smaller "student" model. The student is trained to match the teacher's output distribution (typically via kl-divergence-in-distillation), preserving much of the teacher's capability at lower computational cost. KD is a core concept referenced throughout the wiki, including in discussions of exposure-bias and the on-policy-distillation-survey.

References

  • Hinton, Vinyals, & Dean (2015). "Distilling the Knowledge in a Neural Network." arXiv:1503.02531.