Model-agnostic meta-learning for fast adaptation of deep networks
Citations
10,132 citations
5,333 citations
4,384 citations
2,170 citations
Cites background from "Model-agnostic meta-learning for fa..."
..., 2018), investing more heavily in meta-learning (Wang et al., 2016, 2018a; Finn et al., 2017), and exploring multi-agent learning and interaction as a key catalyst for advanced intelligence (Nowak, 2006; Ohtsuki et al....
[...]
2,077 citations
References
111,197 citations
"Model-agnostic meta-learning for fa..." refers methods in this paper
...When training with MAML, we use one gradient update with K = 10 examples with a fixed step size α = 0.01, and use Adam as the metaoptimizer (Kingma & Ba, 2015)....
[...]
30,843 citations
"Model-agnostic meta-learning for fa..." refers methods in this paper
...Our model follows the same architecture as the embedding function used by Vinyals et al. (2016), which has 4 modules with a 3 × 3 convolutions and 64 filters, followed by batch normalization (Ioffe & Szegedy, 2015), a ReLU nonlinearity, and 2 × 2 max-pooling....
[...]
10,447 citations
7,994 citations
"Model-agnostic meta-learning for fa..." refers background in this paper
...Past work has observed that ReLU neural networks are locally almost linear (Goodfellow et al., 2015), which suggests that second derivatives may be close to zero in most cases, partially explaining the good perforFigure 4....
[...]
...Past work has observed that ReLU neural networks are locally almost linear (Goodfellow et al., 2015), which suggests that second derivatives may be close to zero in most cases, partially explaining the good perfor- Figure 4....
[...]
7,930 citations
"Model-agnostic meta-learning for fa..." refers methods in this paper
...The gradient updates are computed using vanilla policy gradient (REINFORCE) (Williams, 1992), and we use trust-region policy optimization (TRPO) as the meta-optimizer (Schulman et al., 2015)....
[...]