On First-Order Meta-Learning Algorithms
This work addresses the challenge of rapid learning in few-shot scenarios for AI systems, presenting incremental improvements with new algorithms like Reptile.
The paper tackles the problem of meta-learning for quick adaptation to new tasks by analyzing first-order algorithms that learn a parameter initialization, showing they perform well on few-shot classification benchmarks and providing theoretical insights into their effectiveness.
This paper considers meta-learning problems, where there is a distribution of tasks, and we would like to obtain an agent that performs well (i.e., learns quickly) when presented with a previously unseen task sampled from this distribution. We analyze a family of algorithms for learning a parameter initialization that can be fine-tuned quickly on a new task, using only first-order derivatives for the meta-learning updates. This family includes and generalizes first-order MAML, an approximation to MAML obtained by ignoring second-order derivatives. It also includes Reptile, a new algorithm that we introduce here, which works by repeatedly sampling a task, training on it, and moving the initialization towards the trained weights on that task. We expand on the results from Finn et al. showing that first-order meta-learning algorithms perform well on some well-established benchmarks for few-shot classification, and we provide theoretical analysis aimed at understanding why these algorithms work.