Abstract: Recent years have seen rapid progress in meta-learning methods, which transfer knowledge across tasks and domains to learn new tasks more efficiently, optimize the learning process itself, and even generate new learning methods from scratch. Meta-learning can be seen as the logical conclusion of the arc that machine learning has undergone in the last decade, from learning classifiers and policies over hand-crafted features, to learning representations over which classifiers and policies operate, and finally to learning algorithms that themselves acquire representations, classifiers, and policies.

Meta-learning methods are of substantial practical interest. For instance, they have been shown to yield new state-of-the-art automated machine learning algorithms and architectures, and have substantially improved few-shot learning systems. Moreover, the ability to improve one’s own learning capabilities through experience can also be viewed as a hallmark of intelligent beings, and there are strong connections with work on human learning in cognitive science and reward learning in neuroscience.