Fast Context Adaptation via Meta-Learning
This work addresses meta-learning challenges for researchers and practitioners, offering a more interpretable and parallelizable method, though it is incremental as it builds directly on MAML.
The paper tackles the problem of meta-overfitting and inefficiency in meta-learning by proposing CAVIA, an extension to MAML that partitions parameters into context and shared parts, leading to improved performance in regression, classification, and reinforcement learning tasks.
We propose CAVIA for meta-learning, a simple extension to MAML that is less prone to meta-overfitting, easier to parallelise, and more interpretable. CAVIA partitions the model parameters into two parts: context parameters that serve as additional input to the model and are adapted on individual tasks, and shared parameters that are meta-trained and shared across tasks. At test time, only the context parameters are updated, leading to a low-dimensional task representation. We show empirically that CAVIA outperforms MAML for regression, classification, and reinforcement learning. Our experiments also highlight weaknesses in current benchmarks, in that the amount of adaptation needed in some cases is small.