Running Pathfinder on Turing.jl models

This tutorial demonstrates how Turing can be used with Pathfinder.

We'll demonstrate with a regression example.

using AbstractMCMC, AdvancedHMC, DynamicPPL, FlexiChains, Pathfinder, Random, Turing
Random.seed!(39)

@model function regress(x)
    α ~ Normal()
    β ~ Normal()
    σ ~ truncated(Normal(); lower=0)
    μ = α .+ β .* x
    y ~ product_distribution(Normal.(μ, σ))
end
x = 0:0.1:10
true_params = (; α=1.5, β=2, σ=2)
# simulate data
y = rand(regress(x) | true_params)[@varname(y)]

model = regress(x) | (; y)
n_chains = 8

For convenience, pathfinder and multipathfinder can take Turing models as inputs and produce MCMCChains.Chains or FlexiChains.VNChain objects as outputs. To access this, we run Pathfinder normally; the chains representation of the draws (defaulting to Chains) is stored in draws_transformed.

result_single = pathfinder(model; ndraws=1_000)
Single-path Pathfinder result
  tries: 1
  draws: 1000
  fit iteration: 14 (total: 14)
  fit ELBO: -213.66 ± 0.09
  fit distribution: MvNormal{Float64, Pathfinder.WoodburyPDMat{Float64, LinearAlgebra.Diagonal{Float64, Vector{Float64}}, Matrix{Float64}, Matrix{Float64}, Pathfinder.WoodburyPDFactorization{Float64, LinearAlgebra.Diagonal{Float64, Vector{Float64}}, LinearAlgebra.QRCompactWYQ{Float64, Matrix{Float64}, Matrix{Float64}}, LinearAlgebra.UpperTriangular{Float64, Matrix{Float64}}}}, Vector{Float64}}(
dim: 3
μ: [1.650897108575027, 1.9311753921107153, 0.5801261338729242]
Σ: [0.11087806908108015 -0.01652721016434848 -0.0014208266829752382; -0.016527210164348445 0.003407129429651901 0.0001943337674883388; -0.0014208266829752417 0.000194333767488339 0.00471830003589957]
)
result_single.draws_transformed
Chains MCMC chain (1000×6×1 Array{Float64, 3}):

Iterations        = 1:1:1000
Number of chains  = 1
Samples per chain = 1000
parameters        = α, β, σ
internals         = logprior, loglikelihood, logjoint

Use `describe(chains)` for summary statistics and quantiles.

To request a different chain type (e.g. VNChain), we can specify the chain_type directly.

pathfinder(model; ndraws=1_000, chain_type=VNChain).draws_transformed
FlexiChain (1000 iterations, 1 chain)
↓ iter=1:1000 | → chain=1:1

Parameter type   VarName
Parameters       α, β, σ
Extra keys       :logprior, :loglikelihood, :logjoint

Note that while Turing's sample methods default to initializing parameters from the prior with InitFromPrior, Pathfinder defaults to uniformly sampling them in the range [-2, 2] in unconstrained space (equivalent to Turing's InitFromUniform(-2, 2)). To use Turing's default in Pathfinder, specify init_sampler=InitFromPrior().

result_multi = multipathfinder(model, 1_000; nruns=n_chains, init_sampler=InitFromPrior())
Multi-path Pathfinder result
  runs: 8
  draws: 1000
  Pareto shape diagnostic: 0.25 (good)

The Pareto shape diagnostic indicates that it is likely safe to use these draws to compute posterior estimates.

chns_pf = result_multi.draws_transformed
describe(chns_pf)
Chains MCMC chain (1000×6×1 Array{Float64, 3}):

Iterations        = 1:1:1000
Number of chains  = 1
Samples per chain = 1000
parameters        = α, β, σ
internals         = logprior, loglikelihood, logjoint

Summary Statistics

  parameters      mean       std      mcse   ess_bulk   ess_tail      rhat   ess_per_sec 
      Symbol   Float64   Float64   Float64    Float64    Float64   Float64       Missing 

           α    1.6679    0.3343    0.0108   958.6483   993.2858    1.0026       missing
           β    1.9288    0.0582    0.0019   919.4163   901.7198    1.0030       missing
           σ    1.8103    0.1206    0.0039   937.8351   942.3972    0.9996       missing


Quantiles

  parameters      2.5%     25.0%     50.0%     75.0%     97.5% 
      Symbol   Float64   Float64   Float64   Float64   Float64 

           α    0.9963    1.4330    1.6855    1.8904    2.3074
           β    1.8174    1.8899    1.9284    1.9616    2.0491
           σ    1.5897    1.7283    1.8111    1.8906    2.0518

We can also use these draws to initialize MCMC sampling with InitFromParams.

params = AbstractMCMC.to_samples(DynamicPPL.ParamsWithStats, chns_pf[1:n_chains, :, :], model)
initial_params = [InitFromParams(p.params) for p in vec(params)]
chns = sample(model, Turing.NUTS(), MCMCThreads(), 1_000, n_chains; initial_params, progress=false)
describe(chns)
┌ Warning: Only a single thread available: MCMC chains are not sampled in parallel
@ AbstractMCMC ~/.julia/packages/AbstractMCMC/oqm6Y/src/sample.jl:544
┌ Info: Found initial step size
  ϵ = 0.025
┌ Info: Found initial step size
  ϵ = 0.0484375
┌ Info: Found initial step size
  ϵ = 0.025
┌ Info: Found initial step size
  ϵ = 0.0234375
┌ Info: Found initial step size
  ϵ = 0.049218750000000006
┌ Info: Found initial step size
  ϵ = 0.2
┌ Info: Found initial step size
  ϵ = 0.025
┌ Info: Found initial step size
  ϵ = 0.05
Chains MCMC chain (1000×17×8 Array{Float64, 3}):

Iterations        = 501:1:1500
Number of chains  = 8
Samples per chain = 1000
Wall duration     = 4.08 seconds
Compute duration  = 2.72 seconds
parameters        = α, β, σ
internals         = n_steps, is_accept, acceptance_rate, log_density, hamiltonian_energy, hamiltonian_energy_error, max_hamiltonian_energy_error, tree_depth, numerical_error, step_size, nom_step_size, logprior, loglikelihood, logjoint

Summary Statistics

  parameters      mean       std      mcse    ess_bulk    ess_tail      rhat   ess_per_sec 
      Symbol   Float64   Float64   Float64     Float64     Float64   Float64       Float64 

           α    1.6506    0.3406    0.0060   3180.4412   3591.6585    1.0028     1167.9916
           β    1.9311    0.0605    0.0010   3313.3507   3921.0353    1.0018     1216.8016
           σ    1.8163    0.1250    0.0018   4597.5509   4806.0162    1.0017     1688.4138


Quantiles

  parameters      2.5%     25.0%     50.0%     75.0%     97.5% 
      Symbol   Float64   Float64   Float64   Float64   Float64 

           α    0.9812    1.4247    1.6505    1.8758    2.3321
           β    1.8109    1.8912    1.9312    1.9722    2.0502
           σ    1.5917    1.7288    1.8096    1.8951    2.0835

We can use Pathfinder's estimate of the metric and only perform enough warm-up to tune the step size.

inv_metric = result_multi.pathfinder_results[1].fit_distribution.Σ
metric = Pathfinder.RankUpdateEuclideanMetric(inv_metric)
kernel = HMCKernel(Trajectory{MultinomialTS}(Leapfrog(0.0), GeneralisedNoUTurn()))
adaptor = StepSizeAdaptor(0.8, 1.0)  # adapt only the step size
nuts = AdvancedHMC.HMCSampler(kernel, metric, adaptor)

n_adapts = 50
n_draws = 1_000
chns = sample(
    model,
    externalsampler(nuts),
    MCMCThreads(),
    n_draws + n_adapts,
    n_chains;
    n_adapts,
    initial_params,
    progress=false,
)[n_adapts + 1:end, :, :]  # drop warm-up draws
describe(chns)
┌ Warning: Only a single thread available: MCMC chains are not sampled in parallel
@ AbstractMCMC ~/.julia/packages/AbstractMCMC/oqm6Y/src/sample.jl:544
[ Info: Found initial step size 1.6
[ Info: Found initial step size 1.6500000000000001
[ Info: Found initial step size 1.6
[ Info: Found initial step size 3.2
[ Info: Found initial step size 1.6
[ Info: Found initial step size 1.6
[ Info: Found initial step size 1.6
[ Info: Found initial step size 1.6
Chains MCMC chain (1000×17×8 Array{Float64, 3}):

Iterations        = 51:1:1050
Number of chains  = 8
Samples per chain = 1000
Wall duration     = 1.56 seconds
Compute duration  = 1.39 seconds
parameters        = α, β, σ
internals         = n_steps, is_accept, acceptance_rate, log_density, hamiltonian_energy, hamiltonian_energy_error, max_hamiltonian_energy_error, tree_depth, numerical_error, step_size, nom_step_size, logprior, loglikelihood, logjoint

Summary Statistics

  parameters      mean       std      mcse    ess_bulk    ess_tail      rhat   ess_per_sec 
      Symbol   Float64   Float64   Float64     Float64     Float64   Float64       Float64 

           α    1.6426    0.3416    0.0039   7880.5909   6002.8792    1.0003     5665.4140
           β    1.9322    0.0595    0.0007   7963.7002   6011.4621    1.0005     5725.1619
           σ    1.8139    0.1279    0.0016   6991.9770   5346.5114    1.0007     5026.5831


Quantiles

  parameters      2.5%     25.0%     50.0%     75.0%     97.5% 
      Symbol   Float64   Float64   Float64   Float64   Float64 

           α    0.9726    1.4160    1.6391    1.8725    2.3134
           β    1.8172    1.8920    1.9323    1.9721    2.0487
           σ    1.5866    1.7261    1.8068    1.8917    2.0891

See Initializing HMC with Pathfinder for further examples.