Replaces #1437 - additional documentation for filldist and arraydist #1751
Conversation
|
Could this PR get a more informative description? |
| k = length(unique(g)) | ||
| a ~ filldist(Exponential(), k) # = Product(fill(Exponential(), k)) | ||
| mu = a[g] | ||
| x .~ Normal.(mu) |
There was a problem hiding this comment.
Wouldn't an MvNormal be more performant here?
Codecov Report
@@ Coverage Diff @@
## master #1751 +/- ##
===========================================
+ Coverage 65.56% 81.16% +15.60%
===========================================
Files 25 24 -1
Lines 1661 1476 -185
===========================================
+ Hits 1089 1198 +109
+ Misses 572 278 -294
Continue to review full report at Codecov.
|
Pull Request Test Coverage Report for Build 1587251083
💛 - Coveralls |
Co-authored-by: Rik Huijzer <[email protected]>
trappmartin
changed the title
| @@ -545,6 +545,41 @@ The `Gibbs` sampler can be used to specify unique automatic differentation backe | |||
|
|
|||
| For more details of compositional sampling in Turing.jl, please check the corresponding [paper](http://proceedings.mlr.press/v84/ge18b.html). | |||
|
|
|||
| ### Working with filldist and arraydist | |||
|
|
|||
| Turing provides `filldist(dist::Distribution, n::Int)` and `arraydist(dists::AbstractVector{<:Distribution})` as a simplified interface to construct product distributions, e.g., to model a set of variables that share the same structure but vary by group. | |||
There was a problem hiding this comment.
Just a heads up, rather sooner than later they will be integrated and generalized in Distributions: JuliaStats/Distributions.jl#1391 I am working on a PR to DistributionsAD that would fix the errors mentioned in the last comments; then we can switch product_distribution completely from Product to ProductDistribution.
There was a problem hiding this comment.
Lovely improvements - I look forward to the new Distributions release!
|
many thanks, @trappmartin! |
This PR is aimed to replace #1437.