Early exiting using confidence measures

Introduced by Teerapittayanon et al. in BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks

Exit whenever the model is confident enough allowing early exiting from hidden layers

Source: BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks

Paper	Code	Results	Date	Stars

Task	Papers	Share
Image Classification	4	6.06%
Language Modelling	4	6.06%
Computational Efficiency	3	4.55%
Model Compression	3	4.55%
Text Generation	3	4.55%
Text Classification	3	4.55%
Edge-computing	3	4.55%
Natural Language Understanding	2	3.03%
Named Entity Recognition (NER)	2	3.03%

This feature is experimental; we are continuously improving our matching algorithm.

Component	Type	Add Remove
🤖 No Components Found	You can add them if they exist; e.g. Mask R-CNN uses RoIAlign