Meta Learning

In this work we develop a gradient-based meta-learning algorithm for efficient, online continual learning, that is robust and scalable to real-world visual benchmarks.