Variational Inference in Gaussian Processes for non-linear time series
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Tim Hughes.
Modelling non-linear dynamical systems from observations (system identification) is often a pre-requisite for designing good controllers. Despite the fundamental importance of this problem, good models and inference methods remain a technical challenge. The difficulties include having to deal with noisy and incomplete measurements and treating suitably flexible non-linear models. Sufficiently flexible models generally won’t have succinct representations, and large numbers of free parameters requires principled approaches to issues such as overfitting. In this talk I will present a couple of recent developments in machine learning which together allows approximate fully Bayesian inference jointly over both latent states and non-linear transition models. This elegant and practical framework relies on variational inference in non-linear Gaussian process state space models.
This talk is part of the CUED Control Group Seminars series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|