Fast Adaptation in Generative Models with Generative Matching Networks

Sergey Bartunov, Dmitry P. Vetrov

TL;DR: A nonparametric conditional VAE suitable for few-shot learning

Abstract:

We develop a new class of deep generative model called generative matching networks (GMNs) which is inspired by the recently proposed matching networks for one-shot learning in discriminative tasks.
By conditioning on the additional input dataset, generative matching networks may instantly learn new concepts that were not available during the training but conform to a similar generative process, without explicit limitations on the number of additional input objects or the number of concepts they represent.
Our experiments on the Omniglot dataset demonstrate that GMNs can significantly improve predictive performance on the fly as more additional data is available and generate examples of previously unseen handwritten characters once only a few images of them are provided.

Keywords: Deep learning, Unsupervised Learning

Conflicts: hse.ru, msu.ru, skoltech.ru, google.com, company.yandex.ru

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:1612.02192/code)