Generalized Inner Loop Meta-Learning

By Edward Grefenstette, Brandon Amos, Denis Yarats, Phu Mon Htut, Artem Molchanov, F. Meier, Douwe Kiela, Kyunghyun Cho, Soumith Chintala. 2019

169 citations recorded by the source corpus.

Many (but not all) approaches self-qualifying as "meta-learning" in deep learning and reinforcement learning fit a common pattern of approximating the solution to a nested optimization problem. In this paper, we give a formalization of this shared pattern, which we call GIMLI, prove its general requirements, and derive a general-purpose algorithm for implementing similar approaches. Based on this analysis and algorithm, we describe a library of our design, higher, which we share with the community to assist and enable future research into these kinds of meta-learning approaches. We end the paper by showcasing the practical applications of this framework and library through illustrative experiments and ablation studies which they facilitate.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Meta-Learning in Neural Networks: A Survey (2020, neutral): “loopmeta-learning framework has been proposed [80].”
Optimizing Millions of Hyperparameters by Implicit Differentiation (2019, background): “Hyperpa-rameter optimization (HO) has a rich history [1, 2], and achieved recent success in scaling due to gradient-based optimizers [3–11].”
Meta Learning via Learned Loss (2019, methodology): “3Our framework is implemented using open-source libraries Higher [9] for convenient second-order derivative computations and Hydra [34] for simplified handling of experiment configurations.”
DReCa: A General Task Augmentation Strategy for Few-Shot Natural Language Inference (2021, methodology): “We implement MAML in Py-torch using the higher library (Grefenstette et al., 2019).”
Automatic Data Augmentation for Generalization in Deep Reinforcement Learning (2020, methodology): “…approaches for automatically selecting an appropriate data augmentation for each task: UCB-DrAC which uses UCB [1] to select an augmentation from a given set, RL2-DrAC which uses RL 2 [55] to do this, and Meta-DrAC which meta-learns [18] the weights of a convolutional network using higher [20].”
Automatic Data Augmentation for Generalization in Reinforcement Learning (2021, methodology): “Meta-DrAC is implemented using the higher library [22].”
Balanced Meta-Softmax for Long-Tailed Visual Recognition (2020, methodology): “[8] Edward Grefenstette, Brandon Amos, Denis Yarats, Phu Mon Htut, Artem Molchanov, Franziska Meier, Douwe Kiela, Kyunghyun Cho, and Soumith Chintala.”
Deep Transformers with Latent Depth (2020, methodology): “This training strategy is inspired by the Generalized Inner Loop Meta-learning [10].”
On Episodes, Prototypical Networks, and Few-shot Learning (2020, background): “This popularity has reached such a point that an “episodic” data-loader is often at the core of new FSL algorithms, a practice facilitated by frameworks such as Deleu et al. (2019) and Grefenstette et al. (2019).”