Models and inference for temporal Gaussian processes
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Dr R.E. Turner.
Reformulating Gaussian process (GP) models as stochastic differential
equations (SDE) opens up new opportunities to draw comparisons between classical signal processing algorithms and the state of the art
inference methods that dominate the current machine learning literature.
In this talk, I’ll discuss the SDE form of the popular “spectral
mixture” GP, showing how in the temporal case we can think of such a
model as performing probabilistic time-frequency analysis. Following
this, I will discuss recent advances in approximate inference for
temporal GP models, focusing on power expectation propagation (EP),
which is a natural fit for sequential data. Lastly, I will present new
work showing the conditions under which power EP recovers traditional
methods such as the extended Kalman filter.
This talk is part of the Machine Learning @ CUED series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|