SOTAVerified

How to train your MAML

2018-10-22ICLR 2019Code Available1· sign in to hype

Antreas Antoniou, Harrison Edwards, Amos Storkey

arXivPDF

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

The field of few-shot learning has recently seen substantial advancements. Most of these advancements came from casting few-shot learning as a meta-learning problem. Model Agnostic Meta Learning or MAML is currently one of the best approaches for few-shot learning via meta-learning. MAML is simple, elegant and very powerful, however, it has a variety of issues, such as being very sensitive to neural network architectures, often leading to instability during training, requiring arduous hyperparameter searches to stabilize training and achieve high generalization and being very computationally expensive at both training and inference times. In this paper, we propose various modifications to MAML that not only stabilize the system, but also substantially improve the generalization performance, convergence speed and computational overhead of MAML, which we call MAML++.

Tasks

Few-Shot Image ClassificationFew-Shot LearningMeta-Learning

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
Mini-Imagenet 5-way (1-shot)MAML++Accuracy52.4—Unverified
Mini-Imagenet 5-way (5-shot)MAML++Accuracy67.15—Unverified
OMNIGLOT - 1-Shot, 20-wayMAML++Accuracy97.65—Unverified
OMNIGLOT - 1-Shot, 20-wayMAML++Accuracy97.7—Unverified
OMNIGLOT - 1-Shot, 5-wayMAML++Accuracy99.47—Unverified
OMNIGLOT - 5-Shot, 5-wayMAML++Accuracy99.85—Unverified

Reproductions