Meta Specification

Some nodes in ReactiveMP.jl accept optional meta structure that may be used to change or customise the inference procedure or the way node computes outbound messages. As an example GCV node accepts the approxximation method that will be used to approximate non-conjugate relationships between variables in this node. GraphPPL.jl exports @meta macro to specify node-specific meta and contextual information. For example:

General syntax

@meta macro accepts either regular julia function or a single begin ... end block. For example both are valid:

@meta function create_meta(arg1, arg2)
    ...
end

mymeta = @meta begin 
    ...
end

In the first case it returns a function that return meta upon calling, e.g.

@meta function create_meta(flag)
    ...
end

mymeta = create_meta(true)

and in the second case it returns constraints directly.

mymeta = @meta begin 
    ...
end

Options specification

@meta macro accepts optional list of options as a first argument and specified as an array of key = value pairs, e.g.

mymeta = @meta [ warn = false ] begin 
   ...
end

List of available options:

• warn::Bool - enables/disables various warnings with an incompatible model/meta specification

Meta specification

First, lets start with an example:

meta = @meta begin 
    GCV(x, k, w) -> GCVMetadata(GaussHermiteCubature(20))
end

This meta specification indicates, that for every GCV node in the model with x, k and w as connected variables should use the GCVMetadata(GaussHermiteCubature(20)) meta object.

You can have a list of as many as possible meta specification entries for different nodes:

meta = @meta begin 
    GCV(x1, k1, w1) -> GCVMetadata(GaussHermiteCubature(20))
    GCV(x2, k2, w3) -> GCVMetadata(GaussHermiteCubature(30))
    NormalMeanVariance(y, x) -> MyCustomMetaObject(arg1, arg2)
end

To create a model with extra constraints user may pass optional meta positional argument (comes either first, or after constraints if there are any) for the model function:

@model function my_model(arguments...)
   ...
end

constraints = @constraints begin 
    ...
end

meta = @meta begin 
    ...
end

# both are valid
model, (x, y) = my_model(meta, arguments...)
model, (x, y) = my_model(constraints, meta, arguments...)

Alternatively, it is possible to use Model function directly or resort to the automatic inference function that accepts meta keyword argument.