Add column-pivoted QR factorization fallback on failed LU factorization #617
Conversation
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
In action:
```julia
using LinearSolve
A = [1.0 0 0 0
0 1 0 0
0 0 1 0
0 0 0 0]
b = rand(4)
prob = LinearProblem(A, b)
sol = solve(prob, LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback=false))
@test sol.retcode === ReturnCode.Failure
@test sol.u == zeros(4)
sol = solve(prob)
@test sol.u ≈ svd(A)\b
```
Previously it would just fail, now it falls back to the column-pivoted QR in order to successfully factorize when necessary.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Comment on lines
+131
to
+133
| * `safetyfallback`: determines whether to fallback to a column-pivoted QR factorization | ||
| when an LU factorization fails. This can be required if `A` is rank-deficient. Defaults | ||
| to true. |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| * `safetyfallback`: determines whether to fallback to a column-pivoted QR factorization | |
| when an LU factorization fails. This can be required if `A` is rank-deficient. Defaults | |
| to true. | |
| - `safetyfallback`: determines whether to fallback to a column-pivoted QR factorization | |
| when an LU factorization fails. This can be required if `A` is rank-deficient. Defaults | |
| to true. |
| struct DefaultLinearSolver <: SciMLLinearSolveAlgorithm | ||
| alg::DefaultAlgorithmChoice.T | ||
| safetyfallback::Bool | ||
| DefaultLinearSolver(alg; safetyfallback=true) = new(alg,safetyfallback) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| DefaultLinearSolver(alg; safetyfallback=true) = new(alg,safetyfallback) | |
| DefaultLinearSolver(alg; safetyfallback = true) = new(alg, safetyfallback) |
| @@ -41,6 +41,11 @@ end | |||
| ex = Expr(:if, ex.args...) | |||
| end | |||
|
|
|||
| # Handle special case of Column-pivoted QR fallback for LU | |||
| function __setfield!(cache::DefaultLinearSolverInit, alg::DefaultLinearSolver, v::LinearAlgebra.QRPivoted) | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| function __setfield!(cache::DefaultLinearSolverInit, alg::DefaultLinearSolver, v::LinearAlgebra.QRPivoted) | |
| function __setfield!(cache::DefaultLinearSolverInit, | |
| alg::DefaultLinearSolver, v::LinearAlgebra.QRPivoted) |
| sol = SciMLBase.solve!(cache, $(algchoice_to_alg(alg)), args...; kwargs...) | ||
| if sol.retcode === ReturnCode.Failure && alg.safetyfallback | ||
| ## TODO: Add verbosity logging here about using the fallback | ||
| sol = SciMLBase.solve!(cache, QRFactorization(ColumnNorm()), args...; kwargs...) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| sol = SciMLBase.solve!(cache, QRFactorization(ColumnNorm()), args...; kwargs...) | |
| sol = SciMLBase.solve!( | |
| cache, QRFactorization(ColumnNorm()), args...; kwargs...) |
Comment on lines
+165
to
+167
| 0 0 1 0 | ||
| 0 0 0 0] |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| 0 1 0 0 | |
| 0 0 1 0 | |
| 0 0 0 0] | |
| 0 1 0 0 | |
| 0 0 1 0 | |
| 0 0 0 0] |
| 0 0 0 0] | ||
| b = rand(4) | ||
| prob = LinearProblem(A, b) | ||
| sol = solve(prob, LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback=false)) |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| sol = solve(prob, LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback=false)) | |
| sol = solve(prob, | |
| LinearSolve.DefaultLinearSolver( | |
| LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback = false)) |
| @test sol.u == zeros(4) | ||
|
|
||
| sol = solve(prob) | ||
| @test sol.u ≈ svd(A)\b |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| @test sol.u ≈ svd(A)\b | |
| @test sol.u ≈ svd(A) \ b |
ChrisRackauckas
added a commit
to SciML/NonlinearSolve.jl
that referenced
this pull request
May 26, 2025
Requires SciML/LinearSolve.jl#617 This now uses the DefaultLinearSolver sum type in order to do the QR factorization fallback. This removes the `additional_lincache` which was `::Any`. This fixes a core type instability on the `LinearSolveResult` that was causing a pretty decently large slow down to the Newton methods. This also makes it so if the user specifically says they want LU, it don't do anything different.
| LUFactorization(), | ||
| QRFactorization(), | ||
| GenericLUFactorization(), | ||
| LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback=false), |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback=false), | |
| LinearSolve.DefaultLinearSolver( | |
| LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback = false), |
| GenericLUFactorization(), | ||
| LinearSolve.DefaultLinearSolver(LinearSolve.DefaultAlgorithmChoice.LUFactorization; safetyfallback=false), | ||
| RFLUFactorization(), | ||
| NormalCholeskyFactorization(), |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| NormalCholeskyFactorization(), | |
| NormalCholeskyFactorization() |
| sol = solve!(linsolve) | ||
| @test SciMLBase.successful_retcode(sol.retcode) | ||
| end | ||
| end |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| end | |
| end |
ChrisRackauckas
added a commit
to SciML/NonlinearSolve.jl
that referenced
this pull request
May 26, 2025
Requires SciML/LinearSolve.jl#617 This now uses the DefaultLinearSolver sum type in order to do the QR factorization fallback. This removes the `additional_lincache` which was `::Any`. This fixes a core type instability on the `LinearSolveResult` that was causing a pretty decently large slow down to the Newton methods. This also makes it so if the user specifically says they want LU, it don't do anything different.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
In action:
Previously it would just fail, now it falls back to the column-pivoted QR in order to successfully factorize when necessary.