Skip to content

Commit

Permalink
refactoring
Browse files Browse the repository at this point in the history
  • Loading branch information
@mohamed82008
mohamed82008 committed Nov 15, 2023
1 parent f1c78ab commit 3111adc
Show file tree
Hide file tree
Showing 3 changed files with 244 additions and 141 deletions.
3 changes: 2 additions & 1 deletion Project.toml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,7 @@ julia = "1"
[extras]
NonconvexNLopt = "b43a31b8-ff9b-442d-8e31-c163daa8ab75"
Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"

[targets]
test = ["NonconvexNLopt", "Test"]
test = ["NonconvexNLopt", "Test", "Zygote"]
182 changes: 109 additions & 73 deletions src/bayesian.jl
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,29 +4,51 @@ mutable struct ZeroOrderGPSurrogate <: Function
f::Function
gps::Vector{<:GP}
X::Vector{Vector{Float64}}
y::Union{Vector{Vector{Float64}}, Vector{Float64}}
y::Union{Vector{Vector{Float64}},Vector{Float64}}
noise::Float64
std_multiple::Float64
mode::Symbol
N::Int
every::Int
skip::Int
last::Union{Symbol, Int}
last::Union{Symbol,Int}
fit_prior::Bool
fit_noise::Bool
counter::Int
end
function ZeroOrderGPSurrogate(
f, x; kernel = SqExponentialKernel(), X = [x], y = [f(x)], noise = 1e-8,
std_multiple = 3.0, mode = :interval, every = 5, skip = 10, last = :all,
fit_prior = true, fit_noise = false,
f,
x;
kernel = SqExponentialKernel(),
X = [x],
y = [f(x)],
noise = 1e-8,
std_multiple = 3.0,
mode = :interval,
every = 5,
skip = 10,
last = :all,
fit_prior = true,
fit_noise = false,
)
@assert noise > 0
gps = [GP(AbstractGPs.ConstMean(0.0), kernel) for _ in 1:length(y[1])]
gps = [GP(AbstractGPs.ConstMean(0.0), kernel) for _ = 1:length(y[1])]
N = length(y[1])
return ZeroOrderGPSurrogate(
f, gps, X, y, noise, std_multiple, mode, N, every, skip, last,
fit_prior, fit_noise, 0,
f,
gps,
X,
y,
noise,
std_multiple,
mode,
N,
every,
skip,
last,
fit_prior,
fit_noise,
0,
)
end

Expand Down Expand Up @@ -66,25 +88,27 @@ function fit_mle!(s, gps, X, y, noise, last, fit_noise)
_X = X
_y = y
else
_X = X[max(end-last+1, 1):end]
_y = y[max(end-last+1, 1):end]
_X = X[max(end - last + 1, 1):end]
_y = y[max(end - last + 1, 1):end]
end
obj(θ) = begin
if _y isa Vector{<:Vector}
offset = 0
return -sum(map(1:length(_y[1])) do i
l1 = length(xs_uns_1[i][1])
un1 = xs_uns_1[i][2]
l2 = length(xs_lbs_ubs_uns_2[i][1])
un2 = xs_lbs_ubs_uns_2[i][4]
_gp = GP(un1(θ[offset+1:offset+l1]), un2(θ[offset+l1+1:offset+l1+l2]))
offset += l1 + l2
if fit_noise
return logpdf(_gp(_X, θ[end]), getindex.(_y, i))
else
return logpdf(_gp(_X, noise), getindex.(_y, i))
end
end)
return -sum(
map(1:length(_y[1])) do i
l1 = length(xs_uns_1[i][1])
un1 = xs_uns_1[i][2]
l2 = length(xs_lbs_ubs_uns_2[i][1])
un2 = xs_lbs_ubs_uns_2[i][4]
_gp = GP(un1(θ[offset+1:offset+l1]), un2(θ[offset+l1+1:offset+l1+l2]))
offset += l1 + l2
if fit_noise
return logpdf(_gp(_X, θ[end]), getindex.(_y, i))
else
return logpdf(_gp(_X, noise), getindex.(_y, i))
end
end,
)
else
l1 = length(xs_uns_1[1][1])
un1 = xs_uns_1[1][2]
Expand Down Expand Up @@ -112,13 +136,15 @@ function fit_mle!(s, gps, X, y, noise, last, fit_noise)
out = GP(un1(x[offset+1:offset+l1]), un2(x[offset+l1+1:offset+l1+l2]))
offset += l1 + l2
return out
end, fit_noise ? x[end] : noise
end,
fit_noise ? x[end] : noise
s.gps = gps
s.noise = noise
return s
end
function ChainRulesCore.rrule(::typeof(fit_mle!), s, gp, X, y, noise, last, fit_noise)
return fit_mle!(s, gp, X, y, noise, last, fit_noise), _ -> begin
return fit_mle!(s, gp, X, y, noise, last, fit_noise),
_ -> begin
return ntuple(_ -> NoTangent(), Val(8))
end
end
Expand All @@ -132,20 +158,21 @@ function (s::ZeroOrderGPSurrogate)(x)
if s.fit_prior && s.counter > s.skip * s.every && (s.counter % s.every) == 0
fit_mle!(s, s.gps, s.X, s.y, s.noise, s.last, s.fit_noise)
end
return Interval.(y, y)
return Interval.(y, y)
else
if eltype(s.y) <: Real
return _call_gp(s.gps[1], x, s.X, s.y, s.noise, s.std_multiple)
else
return map(i -> _call_gp(s.gps[i], x, s.X, getindex.(s.y, i), s.noise, s.std_multiple), 1:length(s.gps))
return map(
i -> _call_gp(s.gps[i], x, s.X, getindex.(s.y, i), s.noise, s.std_multiple),
1:s.N,
)
end
end
end

function _call_gp(gp, x, X, y, noise, std_multiple)
_m, _v = mean_and_var(posterior(
gp(X, noise), y,
), [x])
_m, _v = mean_and_var(posterior(gp(X, noise), y), [x])
m, v = _m[1], _v[1]
r = std_multiple * sqrt(v)
return Interval(m - r, m + r)
Expand All @@ -163,7 +190,7 @@ get_lo(x::AbstractVector) = map(get_lo, x)
_lower_f(s) = get_lo ∘ s
_equality_f(s) = x -> begin
t = s(x)
return [getproperty.(t, :lo); .- getproperty(t, :hi)]
return [getproperty.(t, :lo); .-getproperty(t, :hi)]
end

function surrogate_model(vecmodel::VecModel; kwargs...)
Expand All @@ -172,11 +199,7 @@ function surrogate_model(vecmodel::VecModel; kwargs...)
if :expensive in vecmodel.objective.flags
s = surrogate(vecmodel.objective.f, copy(x0); kwargs...)
push!(surrogates, s)
obj = Objective(
_lower_f(s),
vecmodel.objective.multiple,
vecmodel.objective.flags,
)
obj = Objective(_lower_f(s), vecmodel.objective.multiple, vecmodel.objective.flags)
else
obj = vecmodel.objective
end
Expand All @@ -188,32 +211,41 @@ function surrogate_model(vecmodel::VecModel; kwargs...)
!(:expensive in c.flags)
end
eq_constraints = VectorOfFunctions(cheap_eq_constraints)
ineq_constraints1 = Tuple(mapreduce(vcat, expensive_eq_constraints; init = Union{}[]) do c
@assert c isa EqConstraint
s = surrogate(c, copy(x0); kwargs...)
push!(surrogates, s)
return IneqConstraint(
_equality_f(s), [zero.(c.rhs); zero.(c.rhs)], c.dim * 2, c.flags,
)
end)
ineq_constraints1 = Tuple(
mapreduce(vcat, expensive_eq_constraints; init = Union{}[]) do c
@assert c isa EqConstraint
s = surrogate(c, copy(x0); kwargs...)
push!(surrogates, s)
return IneqConstraint(
_equality_f(s),
[zero.(c.rhs); zero.(c.rhs)],
c.dim * 2,
c.flags,
)
end,
)
ineq_constraints2 = map(vecmodel.ineq_constraints.fs) do c
@assert c isa IneqConstraint
if :expensive in c.flags
s = surrogate(c, copy(x0); kwargs...)
push!(surrogates, s)
return IneqConstraint(
_lower_f(s), zero.(c.rhs), c.dim, c.flags,
)
return IneqConstraint(_lower_f(s), zero.(c.rhs), c.dim, c.flags)
else
return c
end
end
ineq_constraints = VectorOfFunctions(
(ineq_constraints1..., ineq_constraints2...),
)
ineq_constraints = VectorOfFunctions((ineq_constraints1..., ineq_constraints2...),)
return VecModel(
obj, eq_constraints, ineq_constraints, vecmodel.sd_constraints, vecmodel.box_min, vecmodel.box_max, vecmodel.init, vecmodel.integer,
), Tuple(surrogates)
obj,
eq_constraints,
ineq_constraints,
vecmodel.sd_constraints,
vecmodel.box_min,
vecmodel.box_max,
vecmodel.init,
vecmodel.integer,
),
Tuple(surrogates)
end

function update_surrogates!(model, surrogates, x)
Expand All @@ -229,7 +261,7 @@ function update_surrogates!(model, surrogates, x)
return o, i, e
end

struct BayesOptOptions{S, N <: NamedTuple}
struct BayesOptOptions{S,N<:NamedTuple}
sub_options::S
maxiter::Int
initialize::Bool
Expand Down Expand Up @@ -277,15 +309,21 @@ function BayesOptOptions(;
)
end

mutable struct BayesOptWorkspace{M <: VecModel, S1, X <: AbstractVector, O <: BayesOptOptions, S2 <: Union{Tuple, AbstractVector}} <: Workspace
mutable struct BayesOptWorkspace{
M<:VecModel,
S1,
X<:AbstractVector,
O<:BayesOptOptions,
S2<:Union{Tuple,AbstractVector},
} <: Workspace
model::M
sub_workspace::S1
x0::X
options::O
surrogates::S2
end

struct BayesOptResult{M1, M2, S1 <: AbstractResult, S2} <: AbstractResult
struct BayesOptResult{M1,M2,S1<:AbstractResult,S2} <: AbstractResult
minimizer::M1
minimum::M2
niters::Int
Expand All @@ -308,8 +346,12 @@ Returns an iterator over `N` elements of a Sobol sequence between `lowerbounds`
and `upperbounds`. The first `N` elements of the Sobol sequence are skipped for
better uniformity (see https://github.com/stevengj/Sobol.jl)
"""
function ScaledSobolIterator(lowerbounds, upperbounds, N;
seq = SobolSeq(length(lowerbounds)))
function ScaledSobolIterator(
lowerbounds,
upperbounds,
N;
seq = SobolSeq(length(lowerbounds)),
)
N > 0 && skip(seq, N)
ScaledSobolIterator(lowerbounds, upperbounds, N, seq)
end
Expand Down Expand Up @@ -389,33 +431,27 @@ function optimize!(workspace::BayesOptWorkspace)
s.mode = :interval
end
end
return BayesOptResult(
minimizer, minval, iter, r, surrogates,
)
return BayesOptResult(minimizer, minval, iter, r, surrogates)
end

struct BayesOptAlg{A} <: AbstractOptimizer
sub_alg::A
end

function Workspace(
model::VecModel, optimizer::BayesOptAlg, x0::AbstractVector;
options::BayesOptOptions, surrogates = nothing, kwargs...,
model::VecModel,
optimizer::BayesOptAlg,
x0::AbstractVector;
options::BayesOptOptions,
surrogates = nothing,
kwargs...,
)
if surrogates !== nothing
smodel = model
else
smodel, surrogates = surrogate_model(model; options.nt...)
end
sub_workspace = Workspace(
smodel, optimizer.sub_alg, x0;
options = options.sub_options, kwargs...,
)
return BayesOptWorkspace(
model,
sub_workspace,
x0,
options,
surrogates,
)
sub_workspace =
Workspace(smodel, optimizer.sub_alg, x0; options = options.sub_options, kwargs...)
return BayesOptWorkspace(model, sub_workspace, x0, options, surrogates)
end
Loading

2 comments on commit 3111adc

@mohamed82008
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@JuliaRegistrator register

@JuliaRegistrator
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Registration pull request created: JuliaRegistries/General/95346

Tip: Release Notes

Did you know you can add release notes too? Just add markdown formatted text underneath the comment after the text
"Release notes:" and it will be added to the registry PR, and if TagBot is installed it will also be added to the
release that TagBot creates. i.e.

@JuliaRegistrator register

Release notes:

## Breaking changes

- blah

To add them here just re-invoke and the PR will be updated.

Tagging

After the above pull request is merged, it is recommended that a tag is created on this repository for the registered package version.

This will be done automatically if the Julia TagBot GitHub Action is installed, or can be done manually through the github interface, or via:

git tag -a v0.1.5 -m "<description of version>" 3111adc715c08f16b3f7aead04bcbc440ce8d199
git push origin v0.1.5

Please sign in to comment.