Learning to learn by gradient descent by gradient descent

Abstract

This paper introduces and evaluates the idea described in “Learning to learn by gradient descent by gradient descent”, and reports empirical results that helped shape subsequent work in meta-learning, optimization.