Meta-Learning with Variational Bayes

TL;DR

This paper introduces a variational Bayes-based meta-learning approach that creates fast-adapting generative models in latent space, applicable even with unlabeled data, advancing cognitive flexibility in AI.

Contribution

It proposes a novel variational Bayes framework for generative meta-learning that simplifies updates and enhances adaptability without relying on neural network dependencies.

Findings

Theoretical proof that VB updates are independent of generative neural networks for certain models.

Empirical results demonstrating rapid adaptation of latent-space generative models.

Improved flexibility in learning from unlabeled data in meta-learning settings.

Abstract

The field of meta-learning seeks to improve the ability of today's machine learning systems to adapt efficiently to small amounts of data. Typically this is accomplished by training a system with a parametrized update rule to improve a task-relevant objective based on supervision or a reward function. However, in many domains of practical interest, task data is unlabeled, or reward functions are unavailable. In this paper we introduce a new approach to address the more general problem of generative meta-learning, which we argue is an important prerequisite for obtaining human-level cognitive flexibility in artificial agents, and can benefit many practical applications along the way. Our contribution leverages the AEVB framework and mean-field variational Bayes, and creates fast-adapting latent-space generative models. At the heart of our contribution is a new result, showing that for a broad class of deep generative latent variable models, the relevant VB updates do not depend on any generative neural network. The theoretical merits of our approach are reflected in empirical experiments.