Chad Scherrer

Just some notes here as I think through this... `predict_joint` gives us a `SossMLJPredictor`. For the multinomial example, the fields have types ```julia julia> [p => typeof(getproperty(predictor_joint, p)) for p...

Think I'm getting close... Say you start with from `p=predictor_joint` from [example-linear-regression.jl](https://github.com/cscherrer/SossMLJ.jl/blob/c9cd5d65f71cc3a3725a3490240c3a14494990cb/examples/example-linear-regression.jl). Then we can mess with the predictive distribution ```julia julia> p.pred @model (X, σ, β) begin η =...

Just remembered about https://github.com/baggepinnen/MonteCarloMeasurements.jl/issues/22 Lots of great background there, need to reread it myself :) @DilumAluthge let's go ahead with the release and update as this moves ahead.

We really need a uniform sampling interface in Soss, may be good to do something similar to Turing's. In Soss, `dynamicHMC` calls this: ```julia function dynamicHMC( rng::AbstractRNG, m::JointDistribution, _data, N::Int...

Do we want it as a field for the model?

e.g. this works: ```julia julia> m = @model N begin p ~ Uniform() y ~ Binomial(N,p) end; julia> dynamicHMC(m(N=100), (y=20,), 5) 5-element Array{NamedTuple{(:p,),Tuple{Float64}},1}: (p = 0.18743530744069434,) (p = 0.1786246661840475,) (p...

so maybe in the model specification we have `numSamples`, and then pass that along in this way when the sampler is called?

> Where do they store their "num_epochs" or "num_iterations" or whatever? hmm, not sure. I'd guess in this context you'd generally want early stopping. Hhaven't really seen that used in...

There's some subtlety here, because we need this to be flexible for use with lots of different inference methods. @DilumAluthge what do you think of `SossMLJModel` having fields `hyperparams`, `transform`,...

The difficulty here is that the Soss inference API will need to have some more changes. I think we need to avoid having people writing MLJ code that targets it...