Variational implicit processes
36th International Conference on Machine Learning, ICML 2019
MetadataShow full item record
Ma, C., Li, Y., & Hernández-Lobato, J. (2019). Variational implicit processes. 36th International Conference on Machine Learning, ICML 2019, 2019-June 7464-7482. https://doi.org/10.17863/CAM.42186
We introduce the implicit processes (IPs), a stochastic process that places implicitly defined multivariate distributions over any finite collections of random variables. IPs are therefore highly flexible implicit priors over functions, with examples including data simulators, Bayesian neural networks and non-linear transformations of stochastic processes. A novel and efficient approximate inference algorithm for IPs, namely the variational implicit processes (VIPs), is derived using generalised wake-sleep updates. This method returns simple update equations and allows scalable hyper-parameter learning with stochastic optimization. Experiments show that VIPs return better uncertainty estimates and lower errors over existing inference methods for challenging models such as Bayesian neural networks, and Gaussian processes.
External DOI: https://doi.org/10.17863/CAM.42186
This record's URL: https://www.repository.cam.ac.uk/handle/1810/295114