Model compression technique that transfers learned representations from large, accurate teacher neural networks to compact student models optimised for edge deployment, achieving 20-30x size reduction while retaining 97%+ of accuracy through soft-target training, temperature scaling, and layer-wise representation matching.
Semantic Classification
Content
Knowledge Distillation for Edge (AI-0443) — content pending enrichment.