Skip to content
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

[WIP] Mixture Models #79

Closed
wants to merge 6 commits into from
+22 −0
Closed

[WIP] Mixture Models #79

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
dfadb21
mixture model logpdf using Buffer
sivapvarma May 11, 2020
57c75b4
add zygote to dependencies
sivapvarma May 11, 2020
f5fa905
use map do block, remove zygote dependency
sivapvarma May 11, 2020
2326374
filter out nothing from lp
sivapvarma May 12, 2020
5e1cc34
Merge branch 'master' into mixmod
sivapvarma Aug 22, 2020
62c9dcc
Merge branch 'master' into mixmod
sivapvarma Mar 16, 2021
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
1 change: 1 addition & 0 deletions src/DistributionsAD.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@ include("arraydist.jl")
include("filldist.jl")
include("univariate.jl")
include("multivariate.jl")
include("mixturemodels.jl")
include("mvcategorical.jl")
include("matrixvariate.jl")
include("flatten.jl")
Expand Down
21 changes: 21 additions & 0 deletions src/mixturemodels.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
function _mixlogpdf1(d::AbstractMixtureModel, x)
# using the formula below for numerical stability
#
# logpdf(d, x) = log(sum_i pri[i] * pdf(cs[i], x))
# = log(sum_i pri[i] * exp(logpdf(cs[i], x)))
# = log(sum_i exp(logpri[i] + logpdf(cs[i], x)))

Comment on lines +2 to +7
Copy link
Member

Choose a reason for hiding this comment

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

The comment seems a bit useless now that logsumexp is used instead of the manual implementation of the logsumexp trick.

Suggested change
# using the formula below for numerical stability
#
# logpdf(d, x) = log(sum_i pri[i] * pdf(cs[i], x))
# = log(sum_i pri[i] * exp(logpdf(cs[i], x)))
# = log(sum_i exp(logpri[i] + logpdf(cs[i], x)))

pri = probs(d)
indices = findall(!iszero, pri)
lp = map(indices) do i
return logpdf(component(d, i), x) + log(pri[i])
end

return logsumexp(lp)
end



Distributions.logpdf(d::UnivariateMixture{Continuous}, x::Real) = _mixlogpdf1(d, x)
Distributions.logpdf(d::UnivariateMixture{Discrete}, x::Int) = _mixlogpdf1(d, x)
Distributions._logpdf(d::MultivariateMixture, x::AbstractVector) = _mixlogpdf1(d, x)