A Concise Review of Recent Few-shot Meta-learning Methods

few-shot meta-learning的预期是希望model能像人一样，在prior knowledge的基础上，快速学会新的concepts（有点举一反三内味）。

The Framework of Few-shot Meta-learning

Notation and definitions

两个dataset

• ${\mathcal {D}}{base} = \lbrace (X{i},Y_{i}),Y_i \in {\mathcal C}{base} \rbrace^{N{base}}_{i=1}$
• ${\mathcal {D}}{noval} = \lbrace (\tilde X{i},\tilde Y_{i}),\tilde Y_i \in {\mathcal C}{noval} \rbrace^{N{noval}}_{i=1}$

其中$\mathcal C_{base}$和$\mathcal {C}_{noval}$是不相交的。