Modeling with Bounded Partition Functions
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Carl Scheffler.
Many probabilistic models for data are well expressed using energy functions. Typically the (negative) energy is pushed through the exponential function to find the probability distribution over the data. Inference in such models is frequently difficult, as the partition function involves an intractable sum or integral. I will talk about a trick that I used in the Gaussian process density sampler to help sidestep this problem, and talk about how it could be generalised to other energybased probabilistic models. This trick doesn’t necessarily make things easier – it just changes which aspects of the inference problem are difficult. Nonetheless, I hope it will foster interesting discussion.
This talk is part of the Inference Group series.
This talk is included in these lists:
Note that exdirectory lists are not shown.
