8000 Add data_mask to InjectiveData by phajy · Pull Request #71 · fjebaker/SpectralFitting.jl · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Add data_mask to InjectiveData #71

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 9 commits into from
Nov 24, 2023
Merged

Add data_mask to InjectiveData #71

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
ef6bd66
Attempt to add data_mask to InjectiveData
phajy Nov 22, 2023
c58533d
Update src/datasets/injectivedata.jl
phajy Nov 23, 2023
a0308ec
Update src/datasets/injectivedata.jl
phajy Nov 23, 2023
32b34b5
Objective transformer
phajy Nov 23, 2023
1e7380a
Make objective variance for injective data
phajy Nov 23, 2023
c55049f
Update src/datasets/injectivedata.jl
phajy Nov 23, 2023
d048f16
Add test for fitting a contiguously binned dataset
phajy Nov 24, 2023
2a06faa
Fix dataset fitting bug
phajy Nov 24, 2023
56f0ba4
Fix contiguously binned with masked data test
phajy Nov 24, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 17 additions & 7 deletions src/datasets/injectivedata.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@ struct InjectiveData{V} <: AbstractDataset
domain_variance::Union{Nothing,V}
codomain_variance::Union{Nothing,V}
name::String
data_mask::BitVector
end

function InjectiveData(
Expand All @@ -13,8 +14,9 @@ function InjectiveData(
domain_variance = nothing,
codomain_variance = nothing,
name = "[no-name]",
data_mask = BitVector(fill(true, size(codomain)))
)
InjectiveData(domain, codomain, domain_variance, codomain_variance, name)
InjectiveData(domain, codomain, domain_variance, codomain_variance, name, data_mask)
end

supports_contiguosly_binned(::Type{<:InjectiveData}) = true
Expand All @@ -27,24 +29,32 @@ function make_model_domain(::ContiguouslyBinned, dataset::InjectiveData)
push!(domain, domain[end] + Δ)
domain
end
make_objective(::ContiguouslyBinned, dataset::InjectiveData) = dataset.codomain
make_objective(::ContiguouslyBinned, dataset::InjectiveData) = dataset.codomain[dataset.data_mask]

make_model_domain(::OneToOne, dataset::InjectiveData) = dataset.domain
make_objective(::OneToOne, dataset::InjectiveData) = dataset.codomain
make_objective(::OneToOne, dataset::InjectiveData) = dataset.codomain[dataset.data_mask]

function make_objective_variance(
::AbstractDataLayout,
dataset::InjectiveData{V},
)::V where {V}
if !isnothing(dataset.domain_variance)
dataset.codomain_variance
if !isnothing(dataset.codomain_variance)
dataset.codomain_variance[dataset.data_mask]
else
# todo: i dunno just something
1e-8 .* dataset.codomain
1e-8 .* dataset.codomain[dataset.data_mask]
end
end

objective_transformer(::AbstractDataLayout, dataset::InjectiveData) = _DEFAULT_TRANSFORMER()
function objective_transformer(::AbstractDataLayout, dataset::InjectiveData)
function _transformer!!(domain, objective)
@views objective[dataset.data_mask]
end
function _transformer!!(output, domain, objective)
@. output = objective[dataset.data_mask]
end
_transformer!!
end

make_label(dataset::InjectiveData) = dataset.name

Expand Down
17 changes: 17 additions & 0 deletions test/fitting/test-fit-simple-dataset.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,3 +44,20 @@ prob = FittingProblem(model => data)

result = fit(prob, LevenbergMarquadt())
@test result.u[2] ≈ 6.1 atol = 0.1

# fitting a contiguously binned dataset with some masked bins
x = 10 .^ collect(range(-1, 2, 10))
y = x .^ -2.0
y_err = 0.1 .* y
# introduce some bogus data points to ignore
y[2:5] .= 2.0
data = InjectiveData(x, y, codomain_variance=y_err)
# mask out the bogus data points
data.data_mask[2:5] .= false
model = XS_PowerLaw(K=FitParam(1.0E-5), a=FitParam(2.0))
prob = FittingProblem(model => data)
@test SpectralFitting.common_support(model, data) isa SpectralFitting.ContiguouslyBinned
result = fit(prob, LevenbergMarquadt())
@test result.u[1] ≈ 2.55 atol = 0.01
@test result.u[2] ≈ 3.0 atol = 0.05
# note best fit photon index, u[2] should be 3 not 2 becuase y contains bin integrated values not the density
0