-
Notifications
You must be signed in to change notification settings - Fork 6
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.
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
RFC: draft PR for adding Laplace approximation and Expectation Propagation #58
Closed
Closed
Changes from 55 commits
Commits
Show all changes
63 commits
Select commit
Hold shift + click to select a range
3b8a075
initial commit of Laplace approximation & EP
st-- c73e9ce
Laplace demo
st-- 93ec795
bugfix
st-- d62b2dc
bugfix demo
st-- 76afc55
WIP: cleanup; pass around dist_y_given_f explicitly
st-- 46009e4
use ForwardDiff elementwise (much faster)
st-- 0c1a329
format
st-- 88410cc
WIP: gradient
st-- 4853b5f
*Very* WIP - but works
st-- 3c91a22
frule bugfix
st-- 513d34d
clean up comments
st-- 33af436
intermediate cleanup (callbacks not tested)
st-- 30b45ac
fix callback support
st-- 09e5e82
Laplace initial tests
st-- 2591555
make argument order consistent
st-- 5d58f5d
more cleanup
st-- 6c4e56d
cleanup
st-- ead526f
cleanup
st-- fe76626
@info -> @debug and ChainRulesTestUtil workaround
st-- f7439d6
chainrule tests
st-- d71ca20
add @info for res_cold/res_warm
st-- ecb29d9
cleanup
st-- b275332
explicit FiniteDifferences gradient test on laplace_lml
st-- 8135c40
Merge branch 'master' of github.com:JuliaGaussianProcesses/Approximat…
st-- 0d8eb02
format
st-- af07036
format
st-- 6f08e73
remove Zygote dependency - part 1
st-- f84ed86
remove Zygote dependency - part 2
st-- e4aab7e
pkg bugfix
st-- d19003f
update example manifests
st-- 2d43113
fix chainrule test by evaluating frule/rrule on newton_inner_loop bas…
st-- bbd24ec
add compat
st-- ac6486d
clean up test
st-- 96957e3
add missing CRC dependency
st-- 5af23fa
remove workaround
st-- 54acbe3
use more of AbstractGPs API
st-- df88117
clean up laplace_steps
st-- 8d20a65
add laplace example
st-- 30ba663
cleanup
st-- 8209442
cleanup2
st-- def61e7
format
st-- 4566c17
remove demo script
st-- 4095df1
bugfix
st-- 91d0b34
update notebook
st-- bdc3618
bugfix 2
st-- 12655f8
bugfiiiix
st-- b1c0f80
also plot mean
st-- 3c97c97
improve plotting
st-- 81ceb93
Apply suggestions from code review
st-- 82d4695
remove `@ref` that does not work
st-- aca90b1
cleanup
st-- 02b528b
make use of closure fields
st-- cbfbc95
Merge branch 'st/laplace_and_ep' of github.com:JuliaGaussianProcesses…
st-- e788f4c
yaf
st-- ca68222
improved type stability
st-- 9144ed3
replace QuadGK with Gauss-Hermite
st-- 771ef3e
Apply suggestions from code review
st-- b52ef55
more type stability cleanup
st-- 4f45ff6
Merge branch 'st/laplace_and_ep' of github.com:JuliaGaussianProcesses…
st-- 4a694d9
fix test
st-- 62840ad
add missing test file
st-- 6f7a5ba
more explanation on the example script
st-- 5445a95
fix test seed
st-- File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -9,9 +9,11 @@ ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4" | |
Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f" | ||
FastGaussQuadrature = "442a2c76-b920-505d-bb47-c5924d526838" | ||
FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b" | ||
ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210" | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Laplace needs derivatives for the inner loop. Hopefully, ForwardDiff is a sufficiently lightweight dependency. |
||
GPLikelihoods = "6031954c-0455-49d7-b3b9-3e1c99afaf40" | ||
KLDivergences = "3c9cd921-3d3f-41e2-830c-e020174918cc" | ||
LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e" | ||
QuadGK = "1fd47b50-473d-5c70-9696-f719f8f3bcdc" | ||
st-- marked this conversation as resolved.
Show resolved
Hide resolved
|
||
Reexport = "189a3867-3050-52da-a836-e630ba90ab69" | ||
SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b" | ||
Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2" | ||
|
@@ -23,8 +25,10 @@ ChainRulesCore = "1" | |
Distributions = "0.25" | ||
FastGaussQuadrature = "0.4" | ||
FillArrays = "0.12" | ||
ForwardDiff = "0.10" | ||
GPLikelihoods = "0.1, 0.2" | ||
KLDivergences = "0.2.1" | ||
QuadGK = "2" | ||
Reexport = "1" | ||
SpecialFunctions = "1" | ||
StatsBase = "0.33" | ||
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Didn't check how it's done in all the other packages, but particularly when putting packages into an org I find it much more inviting to future contributors if it just says "Copyright the people who contributed to it"... let me know your thoughts on this, happy to discuss it more in depth
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think that for the license we should just have "The JuliaGaussianProcesses organization" and for the author field in the
Project.toml
have `JuliaGaussianProcesses and contributors"There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
My inclination has normally been to go with "Person who created the package + the org", so "Ross Viljoen and the JuliaGaussianProcess org", so that the person who did the initial work retains some credit -- in this case, it would be good to keep Ross' name on it imho. Just a suggestion though.
I've just taken a look through the other packages in the org, and it seems like we're pretty inconsistent. AbstractGPs just mentions JuliaGPs, KF mentions Theo and Turing. We should probably agree to a consistent approach and stick to it, but this PR isn't the place for that.
Consequently, could I suggest that we leave the license as-is in this PR, and update it elsewhere once we've made an org-wide decision?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah this pr isn't gonna get merged anyways: ) just used the opportunity to spark the discussion.
I'd prefer just "the org"; I think it makes more sense from an overall point of view to use this package, for example, to bundle SVGP, Laplace, EP - and I believe we incentivise contributions by new people by not giving the feeling that you're just increasing someone else's credit with your contribution.
From my personal point of view, as much as I want to get myself not to care about attribution as much, I can feel it inside myself that I'd rather create github.com/st--/LaplaceGPs.jl or something like that that I can then put my name on than to add a large set of functionality to "someone else's project". Whereas if it's just listed as "the org" then it feels like by contributing more code I deserve more to be part of the org and take my share of the credit from there. Which is why I changed the text in my branch before pushing it.