KD$^{2}$M: A unifying framework for feature knowledge distillation
This work addresses knowledge distillation for neural networks, but it is incremental as it unifies existing methods rather than introducing a new paradigm.
The paper tackles the problem of knowledge distillation by proposing a unifying framework called KD$^{2}$M that formalizes feature distribution matching, and it benchmarks this approach on computer vision datasets while deriving new theoretical results.
Knowledge Distillation (KD) seeks to transfer the knowledge of a teacher, towards a student neural net. This process is often done by matching the networks' predictions (i.e., their output), but, recently several works have proposed to match the distributions of neural nets' activations (i.e., their features), a process known as \emph{distribution matching}. In this paper, we propose an unifying framework, Knowledge Distillation through Distribution Matching (KD$^{2}$M), which formalizes this strategy. Our contributions are threefold. We i) provide an overview of distribution metrics used in distribution matching, ii) benchmark on computer vision datasets, and iii) derive new theoretical results for KD.