Eric Meissner
Eric Meissner
## Description of changes This is a design proposal and code for adding an MCMC sampler to MXFusion. Right now the method just uses a Normal distribution as the proposal...
There are a few different APIs (clone, replicate, etc) that do slightly different things with regard to copying variables, recursively or not, with or without UUID copying, etc. We should...
In a few cases (namely, when building models of repeated layers), I've noticed people doing things like: ```py z = variables[getattr(self.model, 'inducing_inputs_' + str(layer))] ``` which seems a bit unsafe....
It's a bit difficult (and not obvious) how to get the parameters of a Module out of the InferenceParameters. Right now you have to do something like ``` infr.params[m.y.factor.kernel.lengthscale]```. Having...
The Gaussian process modules do a lot of neat things but could use some documentation on exactly what math they're implementing and where it came from.
This may be useful for Inference algorithms to use during automatic gradient chaining.