Parameterless Transductive Feature Re-representation for Few-Shot Learning

                         Wentao Cui 1 Yuhong Guo 1 2

              Abstract               FSL solutions are generally developed in two branches:
     Recent literature in few-shot learning (FSL) has    meta-learning based methods and fine tuning based meth-
     shown that transductive methods often outperform    ods. In both branches, transductive settings have been re-
     their inductive counterparts. However, most  ...