Meta learning

Learn to learn

$\phi$: learnable component

Find a function $F^{\phi}$ that finds $f^{\theta}$

Training task = training data(support set) + Testing data(query set)

• ML

  Within-task learning

  

  $L(\theta) = \sum e^k$ training data

• Meta

Across-task learning

$L(\phi) = \sum l^n$ training task