model.pdf functionality

[Cygnus-A18]

@kratsg Here is the code ill be referencing:

class Model:
    def __init__(self, *, parameterset: ParameterSet, distributions: DistributionSet, domains: DomainSet):
        self.parameters = {}
        self.parameterset = parameterset

        for parameter in parameterset:
            self.parameters[parameter.name] = boundedscalar(parameter.name, domains[parameter.name])

        self.distributions = {}
        G = nx.DiGraph()
        for dist in distributions:
            G.add_node(dist.name, type="distribution")
            for parameter in dist.parameters:
                if not (parameter in G and G.nodes[parameter]["type"] == "distribution"):
                    G.add_node(parameter, type="parameter")
                G.add_edge(parameter, dist.name)

        for node in nx.topological_sort(G):
            if G.nodes[node]['type'] != "distribution":continue
            self.distributions[node] = distributions[node].expression({**self.parameters, **self.distributions})


    def pdf(self):
        ...
    def logpdf(self):
        ...

I am unsure of how exactly the functionality of .pdf should work, should it be returning just a dict of the calculated values? if so I am a bit confused on how to calculate all the needed pdfs in a general way. Let me know if you have any advice on how to continue or what model.pdf should look like, thanks!

[kratsg]

Here it should be something like

def pdf(self, **kwargs):
  return self.distributions["main_pdf"].eval(kwargs)

or similar. Bigger question is how to really identify what that core/main-pdf, and making sure the model has all the necessary information.