Open longouyang opened 8 years ago
Number of samples was added by #654.
To add to this. In RWebPPL, the user can either pass in the inference algorithm options object from R, or just write it in the WebPPL program as we would normally do. When it's passed in from R, RWebPPL is able to attach the meta-data directly to the posterior object (useful, e.g., in the ggmcmc
package). When inference is specified in WebPPL, however, we don't have access to the meta-data. Concretely, what we currently lack is:
If this were accessible in the distributions object, that would be helpful
Outlining some notes from an in-person conversation with Andreas:
viz.heatMap
, we need to know how many samples there were -- it would be nice if this information was available in marginal distributions.Marginal
distribution is supposed to represent a continuous or discrete posterior (this would be helpful for selecting an appropriate type inviz.auto
).