Meta-Amortized Variational Inference and Learning
This addresses the adaptation challenge in probabilistic modeling for machine learning practitioners, offering a novel approach to improve generalization across distributions.
The paper tackles the problem of generative models struggling to adapt to new distributions by introducing a doubly-amortized variational inference method that learns transferable latent representations across related distributions, showing significant improvements of 10-50% on MNIST and 10-35% on NORB in downstream image classification tasks.
Despite the recent success in probabilistic modeling and their applications, generative models trained using traditional inference techniques struggle to adapt to new distributions, even when the target distribution may be closely related to the ones seen during training. In this work, we present a doubly-amortized variational inference procedure as a way to address this challenge. By sharing computation across not only a set of query inputs, but also a set of different, related probabilistic models, we learn transferable latent representations that generalize across several related distributions. In particular, given a set of distributions over images, we find the learned representations to transfer to different data transformations. We empirically demonstrate the effectiveness of our method by introducing the MetaVAE, and show that it significantly outperforms baselines on downstream image classification tasks on MNIST (10-50%) and NORB (10-35%).