We propose an algorithm for meta-learning that is model-agnostic, in the sense that it is compatible with any model trained with gradient descent and applicable to a variety of different learning problems, including classification, regression, and reinforcement learning. The goal of meta-learning is to train a model on a variety of learning tasks, such that it can solve new learning tasks using only a small number of training samples. In our approach, the parameters of the model are explicitly trained such that a small number of gradient steps with a small amount of training data from a new task will produce good generalization performance on that task. In effect, our method trains the model to be easy to fine-tune. We demonstrate that this approach leads to state-of-the-art performance on two few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies.
核心贡献 · Key contributions
提出 MAML，一种与任何基于梯度下降训练的模型兼容的模型无关的元学习算法。 Proposes MAML, a model-agnostic meta-learning algorithm compatible with any gradient-descent-trained model.
在 Omniglot 和 MiniImagenet 上以更少参数实现少样本分类的最先进性能。 Achieves state-of-the-art few-shot classification on Omniglot and MiniImagenet with fewer parameters.
在少样本回归和基于策略梯度的强化学习中展示快速适应能力。 Demonstrates fast adaptation in few-shot regression and reinforcement learning with policy gradients.
表明一阶近似与完整二阶元梯度性能几乎相同。 Shows that first-order approximation performs nearly as well as full second-order meta-gradient.
提供简单的权重初始化，使少量梯度步骤即可快速微调。 Provides a simple weight initialization that enables rapid fine-tuning with few gradient steps.
局限 · Limitations
需要任务分布进行元训练，可能无法获得。 Requires a distribution over tasks for meta-training, which may not be available.
由于二阶梯度计算成本高，尽管一阶近似有所缓解。 Computationally expensive due to second-order gradients, though first-order approximation helps.
可能难以处理与元训练分布差异很大的任务。 May struggle with tasks that are very different from the meta-training distribution.
性能依赖于内循环步长和梯度步数的选择。 Performance depends on the choice of inner-loop step size and number of gradient steps.
未在图像分类之外的大规模或真实少样本学习基准上评估。 Not evaluated on large-scale or real-world few-shot learning benchmarks beyond image classification.
论文章节 · Sections（共 15）
摘要Abstract
1 引言1 Introduction
2 模型无关的元学习2 Model-Agnostic Meta-Learning
2.1 元学习问题设置2.1 Meta-Learning Problem Set-Up
2.2 一种模型无关的元学习算法2.2 A Model-Agnostic Meta-Learning Algorithm
3 MAML 的三种形式3 Species of MAML
3.1 监督回归与分类3.1 Supervised Regression and Classification