First-order ANIL provably learns representations despite overparametrization
This provides theoretical evidence for the representation learning capabilities of meta-learning methods, addressing a gap in understanding for researchers in machine learning theory, though it is incremental as it builds on existing ANIL frameworks.
The paper tackles the problem of understanding whether model-agnostic meta-learning methods like first-order ANIL can learn shared representations despite architectural misspecifications, showing that with a linear two-layer network and infinite tasks, it successfully learns linear shared representations even with overparametrization, resulting in an asymptotically low-rank solution and good adaptation after a single gradient step.
Due to its empirical success in few-shot classification and reinforcement learning, meta-learning has recently received significant interest. Meta-learning methods leverage data from previous tasks to learn a new task in a sample-efficient manner. In particular, model-agnostic methods look for initialization points from which gradient descent quickly adapts to any new task. Although it has been empirically suggested that such methods perform well by learning shared representations during pretraining, there is limited theoretical evidence of such behavior. More importantly, it has not been shown that these methods still learn a shared structure, despite architectural misspecifications. In this direction, this work shows, in the limit of an infinite number of tasks, that first-order ANIL with a linear two-layer network architecture successfully learns linear shared representations. This result even holds with overparametrization; having a width larger than the dimension of the shared representations results in an asymptotically low-rank solution. The learned solution then yields a good adaptation performance on any new task after a single gradient step. Overall, this illustrates how well model-agnostic methods such as first-order ANIL can learn shared representations.