Closed TMats closed 6 years ago
https://arxiv.org/abs/1703.03400
MAML
NIPSのRL workshopで紹介された
site(動画あり)
github
arXivetimes
DL輪読会
タスクに関して敏感なモデルパラメータを見つけることを目指す
動的システム的にいうと,学習プロセスは新しいタスクの損失関数の感度を最大化することとみなせる
アルゴリズム
回帰
分類
RL
MAMLをベースにしたmeta-learning
https://arxiv.org/abs/1703.03400