Probabilistic Model-Agnostic Meta-Learning
CS394D presentation
Guohao Li
lightaime@gmail.com
Finn C, Xu K, Levine S. Probabilistic model-agnostic meta-learning. In Advances in Neural Information Processing Systems 2018 (pp. 9537-9548).
Background: Model-Agnostic Meta-Learning
MAML learns a pretrained model at meta-training time that can adapt fast to new task at meta-test time
Background: Model-Agnostic Meta-Learning
Ambiguity in Few-Shot Learning
Hard to learn a single model to handle
learn a model distribution!
PMAML
By ligh1994
