About this document
Quantization-aware Knowledge Distillation by Adil Mubashir is a document available to read on EtoBox.
The document presents Quantization-aware Knowledge Distillation (QKD), a method designed to enhance the performance of low-precision quantized deep neural networks (DNNs) by integrating quantization and knowledge distillation. QKD operates in three phases: Self-studying for initialization, Co-studying to adapt the teacher network, and Tutoring to transfer knowledge to the student network, resulting in improved accuracy on datasets like ImageNet and CIFAR. The proposed method outperforms existing techniques,
- Author
- Adil Mubashir
- Language
- EN