Meta-Learning Divergences of Variational Inference
This work addresses a key bottleneck in variational inference for practitioners by automating divergence design, though it is incremental as it builds on existing meta-learning and VI methods.
The paper tackled the problem of automating divergence selection in variational inference by proposing a meta-learning algorithm, which outperformed standard VI on tasks like Gaussian mixture approximation and Bayesian neural network regression.
Variational inference (VI) plays an essential role in approximate Bayesian inference due to its computational efficiency and broad applicability. Crucial to the performance of VI is the selection of the associated divergence measure, as VI approximates the intractable distribution by minimizing this divergence. In this paper we propose a meta-learning algorithm to learn the divergence metric suited for the task of interest, automating the design of VI methods. In addition, we learn the initialization of the variational parameters without additional cost when our method is deployed in the few-shot learning scenarios. We demonstrate our approach outperforms standard VI on Gaussian mixture distribution approximation, Bayesian neural network regression, image generation with variational autoencoders and recommender systems with a partial variational autoencoder.