Variational Gaussian Process Dynamical Systems

Abstract

High dimensional time series are endemic in applications of machine learning such as robotics (sensor data), computational biology (gene expression data), vision (video sequences) and graphics (motion capture data). Practical nonlinear probabilistic approaches to this data are required. In this paper we introduce the variational Gaussian process dynamical system. Our work builds on recent variational approximations for Gaussian process latent variable models to allow for nonlinear dimensionality reduction simultaneously with learning a dynamical prior in the latent space. The approach also allows for the appropriate dimensionality of the latent space to be automatically determined. We demonstrate the model on a human motion capture data set and a series of high resolution video sequences.

16 pages, 19 figures

Cited by in corpus (14)

• Variational Inference: A Review for Statisticians
• Bayesian Temporal Factorization for Multidimensional Time Series Prediction
• ODE${}^2$VAE: Deep generative second order ODEs with Bayesian neural networks
• Manifold Relevance Determination
• Metrics for Probabilistic Geometries
• Variational Inference for Uncertainty on the Inputs of Gaussian Process Models
• SAMBA: Safe Model-Based & Active Reinforcement Learning
• Deep Gaussian Processes: A Survey
• Uncertainty Quantification of Darcy Flow through Porous Media using Deep Gaussian Process
• Forecasting COVID-19 Counts At A Single Hospital: A Hierarchical Bayesian Approach
• Expectation Propagation in Gaussian Process Dynamical Systems: Extended Version
• Evolving-Graph Gaussian Processes
• The Dynamical Gaussian Process Latent Variable Model in the Longitudinal Scenario
• A Variant of Gaussian Process Dynamical Systems