PURSUhInT: In Search of Informative Hint Points Based on Layer Clustering for Knowledge Distillation
This work addresses model compression for deep learning practitioners by improving knowledge distillation efficiency, though it is incremental as it builds on existing hint distillation methods.
The paper tackles the problem of selecting optimal hint points for knowledge distillation by proposing a clustering-based method that chooses hint points from teacher model layers based on clustering metrics, achieving superior compression performance compared to state-of-the-art distillation algorithms on CIFAR-100 and ImageNet datasets.
One of the most efficient methods for model compression is hint distillation, where the student model is injected with information (hints) from several different layers of the teacher model. Although the selection of hint points can drastically alter the compression performance, conventional distillation approaches overlook this fact and use the same hint points as in the early studies. Therefore, we propose a clustering based hint selection methodology, where the layers of teacher model are clustered with respect to several metrics and the cluster centers are used as the hint points. Our method is applicable for any student network, once it is applied on a chosen teacher network. The proposed approach is validated in CIFAR-100 and ImageNet datasets, using various teacher-student pairs and numerous hint distillation methods. Our results show that hint points selected by our algorithm results in superior compression performance compared to state-of-the-art knowledge distillation algorithms on the same student models and datasets.