Learning from a Teacher using Unlabeled Data

Abstract

Knowledge distillation is a widely used technique for model compression. We posit that the teacher model used in a distillation setup, captures relationships between classes, that extend beyond the original dataset. We empirically show that a teacher model can transfer this knowledge to a student model even on an {\it out-of-distribution} dataset. Using this approach, we show promising results on MNIST, CIFAR-10, and Caltech-256 datasets using unlabeled image data from different sources. Our results are encouraging and help shed further light from the perspective of understanding knowledge distillation and utilizing unlabeled data to improve model quality.

References in corpus (4)

• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• A Survey of Model Compression and Acceleration for Deep Neural Networks
• The Open Images Dataset V4: Unified image classification, object detection, and visual relationship detection at scale
• Out-distribution training confers robustness to deep neural networks

Cited by in corpus (2)

• Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Better
• Mosaicking to Distill: Knowledge Distillation from Out-of-Domain Data