Scalable Gaussian Process Variational Autoencoders
arXiv:2010.13472
Abstract
Conventional variational autoencoders fail in modeling correlations between data points due to their use of factorized priors. Amortized Gaussian process inference through GP-VAEs has led to significant improvements in this regard, but is still inhibited by the intrinsic complexity of exact GP inference. We improve the scalability of these methods through principled sparse inference approaches. We propose a new scalable GP-VAE model that outperforms existing approaches in terms of runtime and memory footprint, is easy to implement, and allows for joint end-to-end optimization of all components.
Published at AISTATS 2021
References in corpus (15)
- Auto-Encoding Variational Bayes
- Conditional Generative Adversarial Nets
- Stochastic Backpropagation and Approximate Inference in Deep Generative Models
- Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders
- Composing graphical models with neural networks for structured representations and fast inference
- Inference Suboptimality in Variational Autoencoders
- Rates of Convergence for Sparse Variational Gaussian Process Regression
- Generative Models of Visually Grounded Imagination
- EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
- Variational Message Passing with Structured Inference Networks
- Variational Inference for Gaussian Process Models with Linear Complexity
- Scalable Gaussian Processes on Discrete Domains
- tvGP-VAE: Tensor-variate Gaussian Process Prior Variational Autoencoder
- Longitudinal Variational Autoencoder
- Quadruply Stochastic Gaussian Processes