Repository navigation
add unconstrained Poisson FPOP and isotonic Normal FPOP solvers - #24
Open
williamzhang7792 wants to merge 6 commits into
Open
williamzhang7792 wants to merge 6 commits into
williamzhang7792 wants to merge 6 commits into
Conversation
Medium test: PoissonFPOPunconstrained -- unconstrained optimal partitioning with Poisson loss, using a new set_to_unconstrained_min_of method on PiecewisePoissonLossLog. Hard test: NormalFPOPisotonic -- regularized isotonic regression with Normal loss, using NormalLossPiece (inherits from LossPiece base class) and set_to_min_less_of for the non-decreasing constraint. Both solvers use the package's .C() API (no Rcpp). Includes R wrappers, man pages, and testthat suites (45 tests total).
When the unconstrained Fpop solution is already non-decreasing, the isotonic solver should give the same result.
new solvers, quick start, and file list at the top for reviewer orientation
Owner
|
can you please submit another PR which converts CI to github actions, so we can see if the tests pass here? |
Contributor
Author
CI PR is up: #25 should be good to go once merged, then checks will run here automatically. happy to make any changes if needed. |
This branch has not been deployed
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two new FPOP solvers for the GSoC 2026 gfpop tests:
PoissonFPOPunconstrained: unconstrained optimal partitioning (Poisson loss). Collapses the 2-state up/down model to 1 state, replaces monotonicity operators withset_to_unconstrained_min_of. Validated againstSegmentor3IsBack::Segmentor(model=1).NormalFPOPisotonic: regularized isotonic regression (Normal loss). Usesset_to_min_less_offor non-decreasing constraint.NormalLossPieceinherits from a sharedLossPiecebase class alongsidePoissonLossPieceLog. Validated againstisoreg()(penalty=0) andfpop::Fpop().New files:
src/PoissonFPOPunconstrained.{cpp,h},src/NormalFPOPisotonic.{cpp,h},src/LossPiece.h,R/PoissonFPOPunconstrained.R,R/NormalFPOPisotonic.R,man/PoissonFPOPunconstrained.Rd,man/NormalFPOPisotonic.Rd,tests/testthat/test-PoissonFPOPunconstrained.R(28 tests),tests/testthat/test-NormalFPOPisotonic.R(17 tests).Modified:
src/funPieceListLog.{h,cpp}(addedset_to_unconstrained_min_of),src/interface.cpp(registered new C entry points),README.org.Passes
R CMD check --no-manualwith no new warnings.