Fast Adaptation in Generative Models with Generative Matching Networks

By Sergey Bartunov, Dmitry P. Vetrov. 2016

24 citations recorded by the source corpus.

Despite recent advances, the remaining bottlenecks in deep generative models are necessity of extensive training and difficulties with generalization from small number of training examples. We develop a new generative model called Generative Matching Network which is inspired by the recently proposed matching networks for one-shot learning in discriminative tasks. By conditioning on the additional input dataset, our model can instantly learn new concepts that were not available in the training data but conform to a similar generative process. The proposed framework does not explicitly restrict diversity of the conditioning data and also does not require an extensive inference procedure for training or adaptation. Our experiments on the Omniglot dataset demonstrate that Generative Matching Networks significantly improve predictive performance on the fly as more additional data is available and outperform existing state of the art conditional generative models.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Generative Models

In Signal Processing

Citations in the corpus

Few-shot Autoregressive Density Estimation: Towards Learning to Learn Distributions (2017, methodology): “Few-shot density estimation has been studied previously using matching networks (Bartunov & Vetrov, 2016) and variational autoencoders (VAEs).”
Conditional Neural Processes (2018, background): “Matching networks (Vinyals et al., 2016; Bartunov & Vetrov, 2016) are closely related to CNPs.”
Sample Efficient Adaptive Text-to-Speech (2018, background): “In generative modeling, few-shot learning has been addressed from several perspectives, including matching networks (Bartunov and Vetrov, 2017) and variable inference for memory addressing (Bornschein et al., 2017).”
Neural Processes (2018, background): “Another related extension is generative matching networks (GMNs, Bartunov & Vetrov (2016)), where the conditioning input is pre-processed in a way that is similar to the matching networks model.”
Variational Memory Addressing in Generative Models (2017, methodology): “The model in [3] has a clear advantage when many examples from very few character classes are in memory because it was speciﬁcally designed to extract joint statistics from memory before applying the soft-attention readout.”
Meta-Amortized Variational Inference and Learning (2019, methodology): “The focus has been on few-shot density estimation, with approaches ranging from the use of conditioning (Bartunov and Vetrov 2016) to nested optimization (Reed et al.”
The Kanerva Machine: A Generative Distributed Memory (2018, background): “Similarly, Matching Networks (Vinyals et al., 2016; Bartunov & Vetrov, 2016) and the Neural Episodic Controller (Pritzel et al., 2017) directly store embeddings of data.”
The Variational Homoencoder: Learning to learn high capacity generative models from few examples (2018, methodology): “For likelihood evaluation, our most appropriate comparison is with Generative Matching Networks (Bartunov & Vetrov, 2016) as they also model dependencies within a class.”
Attentive Neural Processes (2019, methodology): “Few-shot density estimation using attention has also been explored extensively in numerous works (Rezende et al., 2016; Reed et al., 2017; Bornschein et al., 2017; Bartunov & Vetrov, 2018).”
Few-shot Image Generation with Elastic Weight Consolidation (2020, background): “Some works focus on few-shot density estimation based on matching networks [1], sequential generative models [32], or autoregressive models [31] but they are limited to generating simple patterns and low resolution results.”