ERROR: MethodError: no method matching Int64(::Irrational{:log2π}) #333
Comments
|
That comes from the wrong promotion for Replace this line by this one |
|
That gives julia> MvNormal(2.0, 1.0)
ERROR: MethodError: no method matching MvNormal(::Float64, ::Float64)so I tried Normal(2.0, 1.0)but that's missing something too ERROR: MethodError: no method matching +(::ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}, ::Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}})
For element-wise addition, use broadcasting with dot syntax: scalar .+ array
Closest candidates are:
+(::Any, ::Any, ::Any, ::Any...) at /nix/store/15kh8pp59zbnhxcxh2l66xr1hzly00y9-julia-bin-1.7.1/share/julia/base/operators.jl:655
+(::ChainRulesCore.Tangent{P}, ::P) where P at ~/.julia/packages/ChainRulesCore/BYuIz/src/differential_arithmetic.jl:162
+(::ForwardDiff.Dual{Tx}, ::RoundingMode) where Tx at ~/.julia/packages/ForwardDiff/pDtsf/src/dual.jl:144
...
Stacktrace:
[1] logdensity(ℓ::var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, params::Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}})
@ Main ~/.../dhmc-test.jl:83
[2] (::LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}})(x::Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}})
@ LogDensityProblems ~/.julia/packages/LogDensityProblems/tWBzE/src/AD_ForwardDiff.jl:20
[3] vector_mode_dual_eval!(f::LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}, x::Vector{Float64})
@ ForwardDiff ~/.julia/packages/ForwardDiff/pDtsf/src/apiutils.jl:37
[4] vector_mode_gradient!(result::DiffResults.MutableDiffResult{1, Float64, Tuple{Vector{Float64}}}, f::LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, x::Vector{Float64}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}})
@ ForwardDiff ~/.julia/packages/ForwardDiff/pDtsf/src/gradient.jl:113
[5] gradient!(result::DiffResults.MutableDiffResult{1, Float64, Tuple{Vector{Float64}}}, f::LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, x::Vector{Float64}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}, ::Val{true})
@ ForwardDiff ~/.julia/packages/ForwardDiff/pDtsf/src/gradient.jl:37
[6] gradient!(result::DiffResults.MutableDiffResult{1, Float64, Tuple{Vector{Float64}}}, f::LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, x::Vector{Float64}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}})
@ ForwardDiff ~/.julia/packages/ForwardDiff/pDtsf/src/gradient.jl:35
[7] logdensity_and_gradient(fℓ::LogDensityProblems.ForwardDiffLogDensity{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}}, x::Vector{Float64})
@ LogDensityProblems ~/.julia/packages/LogDensityProblems/tWBzE/src/AD_ForwardDiff.jl:50
[8] evaluate_ℓ(ℓ::LogDensityProblems.ForwardDiffLogDensity{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}}, q::Vector{Float64})
@ DynamicHMC ~/.julia/packages/DynamicHMC/x07K9/src/hamiltonian.jl:195
[9] initialize_warmup_state(rng::Random._GLOBAL_RNG, ℓ::LogDensityProblems.ForwardDiffLogDensity{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}}; q::Vector{Float64}, κ::GaussianKineticEnergy{Diagonal{Float64, Vector{Float64}}, Diagonal{Float64, Vector{Float64}}}, ϵ::Nothing)
@ DynamicHMC ~/.julia/packages/DynamicHMC/x07K9/src/mcmc.jl:125
[10] initialize_warmup_state(rng::Random._GLOBAL_RNG, ℓ::LogDensityProblems.ForwardDiffLogDensity{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}})
@ DynamicHMC ~/.julia/packages/DynamicHMC/x07K9/src/mcmc.jl:125
[11] mcmc_keep_warmup(rng::Random._GLOBAL_RNG, ℓ::LogDensityProblems.ForwardDiffLogDensity{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}}, N::Int64; initialization::Tuple{}, warmup_stages::Tuple{InitialStepsizeSearch, TuningNUTS{Nothing, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Nothing, DualAveraging{Float64}}}, algorithm::DynamicHMC.NUTS{Val{:generalized}}, reporter::NoProgressReport)
@ DynamicHMC ~/.julia/packages/DynamicHMC/x07K9/src/mcmc.jl:496
[12] macro expansion
@ ~/.julia/packages/UnPack/EkESO/src/UnPack.jl:100 [inlined]
[13] mcmc_with_warmup(rng::Random._GLOBAL_RNG, ℓ::LogDensityProblems.ForwardDiffLogDensity{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}, ForwardDiff.GradientConfig{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2, Vector{ForwardDiff.Dual{ForwardDiff.Tag{LogDensityProblems.var"#46#47"{var"#loglikelihood#1"{Vector{Float64}, Vector{Float64}}}, Float64}, Float64, 2}}}}, N::Int64; initialization::Tuple{}, warmup_stages::Tuple{InitialStepsizeSearch, TuningNUTS{Nothing, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Diagonal, DualAveraging{Float64}}, TuningNUTS{Nothing, DualAveraging{Float64}}}, algorithm::DynamicHMC.NUTS{Val{:generalized}}, reporter::NoProgressReport)
@ DynamicHMC ~/.julia/packages/DynamicHMC/x07K9/src/mcmc.jl:547 |
|
Ah right sorry. logprior(params) = logpdf(MvNormal(2.0 * ones(length(params), 1.0 * ones(length(params)), params)or much better logprior(params) = loglikelihood(Normal(2.0, 1.0), params) |
|
That's working now. Thanks. using AbstractGPs
using Distributions
using StatsFuns
using Plots
default(; legend=:outertopright, size=(700, 400))
using Random
Random.seed!(42) # setting the seed for reproducibility of this notebook
x = [
0.8658165855998895,
0.6661700880180962,
0.8049218148148531,
0.7714303440386239,
0.14790478354654835,
0.8666105548197428,
0.007044577166530286,
0.026331737288148638,
0.17188596617099916,
0.8897812990554013,
0.24323574561119998,
0.028590102134105955,
]
y = [
1.5255314337144372,
3.6434202968230003,
3.010885733911661,
3.774442382979625,
3.3687639483798324,
1.5506452040608503,
3.790447985799683,
3.8689707574953,
3.4933565751758713,
1.4284538820635841,
3.8715350915692364,
3.7045949061144983,
]
x_train = x[1:8]
y_train = y[1:8]
x_test = x[9:end]
y_test = y[9:end]
f = GP(Matern52Kernel())
fx = f(x_train, 0.1)
logpdf(fx, y_train)
p_fx = posterior(fx, y_train)
logpdf(p_fx(x_test), y_test)
function gp_loglikelihood(x, y)
function loglikelihood(params)
kernel =
softplus(params[1]) * (Matern52Kernel() ∘ ScaleTransform(softplus(params[2])))
f = GP(kernel)
fx = f(x, 0.1)
return logpdf(fx, y)
end
return loglikelihood
end
const loglik_train = gp_loglikelihood(x_train, y_train)
using DynamicHMC
using LogDensityProblems
using LinearAlgebra
n_samples = 2_000
n_adapts = 1_000
# Log joint density
function LogDensityProblems.logdensity(ℓ::typeof(loglik_train), params)
return ℓ(params) + logprior(params)
end
logprior(params) = loglikelihood(Normal(2.0, 1.0), params)
# The parameter space is two-dimensional
LogDensityProblems.dimension(::typeof(loglik_train)) = 2
# `loglik_train` does not allow to evaluate derivatives of
# the log-likelihood function
function LogDensityProblems.capabilities(::Type{<:typeof(loglik_train)})
return LogDensityProblems.LogDensityOrder{0}()
end
mcmc_with_warmup(
Random.GLOBAL_RNG,
ADgradient(:ForwardDiff, loglik_train),
n_samples;
reporter=NoProgressReport(),
) |
|
Good that you managed to solve your issue! Actually, the main problem with your initial version was that you used a quite old version of Distributions (0.25.14, the latest version is 0.25.71). The error you saw was fixed last November by JuliaStats/Distributions.jl#1429 and is available in Distributions >= 0.25.32. |
|
That's a good point. It looks like I am constrained by TemporalGPs, which has a note in its readme saying
So attempting to use a newer version of Distributions fails with |
|
TemporalGPs also only supports an old version of AbstractGPs. If it is not mandatory for your example (AFAICT it is not needed for the example in the initial issue), then I recommend using a separate environment with only the mandatory packages and in particular without TemporalGPs for running this example. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I'm trying to learn to use this package. I copied the documentation example of DynamicHMC from https://juliagaussianprocesses.github.io/AbstractGPs.jl/dev/examples/0-intro-1d/
it says
Julia 1.7.1
[99985d1d] AbstractGPs v0.3.9
[bbc10e6e] DynamicHMC v3.3.0
[31c24e10] Distributions v0.25.14
[4c63d2b9] StatsFuns v0.9.9
The text was updated successfully, but these errors were encountered: