Meta-Learning with Shared Amortized Variational Inference
This work addresses a specific issue in meta-learning for few-shot learning, offering an incremental improvement over existing methods.
The paper tackles the problem of conditional prior collapse in empirical Bayes meta-learning by proposing a shared amortized variational inference scheme that prevents the prior from collapsing to a Dirac delta function, achieving improved performance on datasets like miniImageNet, CIFAR-FS, and FC100.
We propose a novel amortized variational inference scheme for an empirical Bayes meta-learning model, where model parameters are treated as latent variables. We learn the prior distribution over model parameters conditioned on limited training data using a variational autoencoder approach. Our framework proposes sharing the same amortized inference network between the conditional prior and variational posterior distributions over the model parameters. While the posterior leverages both the labeled support and query data, the conditional prior is based only on the labeled support data. We show that in earlier work, relying on Monte-Carlo approximation, the conditional prior collapses to a Dirac delta function. In contrast, our variational approach prevents this collapse and preserves uncertainty over the model parameters. We evaluate our approach on the miniImageNet, CIFAR-FS and FC100 datasets, and present results demonstrating its advantages over previous work.