dahyun-kang / renet

[ICCV'21] Official PyTorch implementation of Relational Embedding for Few-Shot Classification
https://arxiv.org/abs/2108.09666
MIT License
113 stars 17 forks source link

Inductive or transductive? #6

Closed Fancy-sf closed 2 years ago

Fancy-sf commented 2 years ago

In the n-way k-shot( k>1) setting, the attended features of the support set are computed by summing k attended features, which are influenced by the query set. I wonder whether it is an inductive or transductive setting in few-shot learning.

dahyun-kang commented 2 years ago

Hello,

Our method is an inductive method in the context of few-shot learning. Averaging k-shot support image features is a prevalent way to incorporate given few-shot images.

For more information about transductive setups, refer to these papers.

Liu et al., Learning to propagate labels: transductive propagation network for few-shot learning, ICLR'19. Liu et al., Prototype rectification for few-shot learning, ECCV'20.

Have a great day! :smiley: