Toward Multimodal Model-Agnostic Meta-Learning
arXiv: Learning, Volume abs/1812.07172, 2018.
Gradient-based meta-learners such as MAML are able to learn a meta-prior from similar tasks to adapt to novel tasks from the same distribution with few gradient updates. One important limitation of such frameworks is that they seek a common initialization shared across the entire task distribution, substantially limiting the diversity of ...More
PPT (Upload PPT)