$π$VAE: a stochastic process prior for Bayesian deep learning with MCMC
This work addresses the problem of scalable Bayesian inference for complex function classes, enabling more practical use of stochastic processes in machine learning applications.
The paper tackles the challenge of efficient inference for stochastic processes in Bayesian deep learning by proposing $\\pi$VAE, a variational autoencoder that serves as a continuous stochastic process prior, achieving state-of-the-art performance in tasks like spatial interpolation with improved accuracy and computational efficiency.
Stochastic processes provide a mathematically elegant way model complex data. In theory, they provide flexible priors over function classes that can encode a wide range of interesting assumptions. In practice, however, efficient inference by optimisation or marginalisation is difficult, a problem further exacerbated with big data and high dimensional input spaces. We propose a novel variational autoencoder (VAE) called the prior encoding variational autoencoder ($π$VAE). The $π$VAE is finitely exchangeable and Kolmogorov consistent, and thus is a continuous stochastic process. We use $π$VAE to learn low dimensional embeddings of function classes. We show that our framework can accurately learn expressive function classes such as Gaussian processes, but also properties of functions to enable statistical inference (such as the integral of a log Gaussian process). For popular tasks, such as spatial interpolation, $π$VAE achieves state-of-the-art performance both in terms of accuracy and computational efficiency. Perhaps most usefully, we demonstrate that the low dimensional independently distributed latent space representation learnt provides an elegant and scalable means of performing Bayesian inference for stochastic processes within probabilistic programming languages such as Stan.