Skip to content

metabayesian_MH is not a meanful model #42

@yebai

Description

@yebai

Differentiating through an MCMC sampler as below is an uncommon operation.

inner_m = inner_model(observation, subj_prior_μ, subj_prior_σ)
# Run the inner Bayesian inference
chns = sample(Xoshiro(468), inner_m, inner_sampler, inner_n_samples, progress = false)
# Extract subjective point estimate
subj_mean_expectationₜ = mean(chns[:mean])

I don't think it is a meaningful example for AD tests since most AD backends won't have a rule for Turing.sample. Meanwhile, automatically derived rules for Turing.sample are meaningless for most MCMC samplers, including MH.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions