Update Optimization requirement from 3 to 3, 5.9

[dependabot]

Updates the requirements on Optimization to permit the latest version.

Release notes

Sourced from Optimization's releases.

v5.9.1

Optimization v5.9.1

Diff since v5.9.0

• Raise the OptimizationBase floor to 5.6.0, which routes solve through the problem's problem_type when wrapping solutions (#1354, #1356)

Non-breaking patch release. Requires SciMLBase 3.54 transitively for the problem_type slot on OptimizationProblem.

---

🤖 Posted by an AI agent — harness: Claude Code, model: claude-opus-5[1m]. Chris asked for this release.

https://claude.ai/code/session_014FEzNTLFutCmTEAZ3zBg5R

Merged pull requests:

• Fix OptimizationManopt reporting MaxIters even when converged (#1258) (@​SebastianM-C)
• Add ConvexOptimization subpackage (disciplined convex programming backend) (#1263) (@​ChrisRackauckas-Claude)
• qa: use the shared reexport documentation policy (#1324) (@​ChrisRackauckas)
• Test OptimizationBBO ZDT1 solution semantically (#1330) (@​ChrisRackauckas-Claude)
• Cap AutoEnzyme Hessian batches at width 8 (#1331) (@​ChrisRackauckas-Claude)
• Release OptimizationBase 5.5.2 (#1332) (@​ChrisRackauckas)
• Fix downgrade compat floors for Ipopt, NLopt, and NOMAD sublibraries (#1333) (@​ChrisRackauckas-Claude)
• Fix OptimizationZygoteExt formatting (#1334) (@​ChrisRackauckas-Claude)
• Release 5.5.3 (#1335) (@​ChrisRackauckas)
• Document restored optimizer reexports (#1336) (@​ChrisRackauckas-Claude)
• Re-enable the SPD manifold test in OptimizationBase (#1339) (@​SebastianM-C)
• Scope Optimization libraries per downstream package (#1341) (@​ChrisRackauckas-Claude)
• Release OptimizationNLopt 0.3.18 (#1342) (@​ChrisRackauckas)
• Release OptimizationOptimJL 0.4.21 (#1343) (@​ChrisRackauckas)
• Release OptimizationOptimisers 0.3.24 (#1344) (@​ChrisRackauckas)
• Release OptimizationManopt 1.6.1 (#1345) (@​ChrisRackauckas)
• Add 32-bit Core CI lane via test_groups.toml (#1346) (@​ChrisRackauckas-Claude)
• ConvexOptimization: lower norm/exp/log objectives through their epigraph (#1347) (@​ChrisRackauckas-Claude)
• ConvexOptimization: lower l1/linf norm and exp/log objectives (#1348) (@​ChrisRackauckas-Claude)
• Add AirspeedVelocity benchmarking CI (#1350) (@​ChrisRackauckas-Claude)
• ConvexOptimization: disciplined parametrized programming (DPP) and fast re-solve (#1351) (@​ChrisRackauckas-Claude)
• Route solve through the problem's problem_type when wrapping solutions (#1354) (@​ChrisRackauckas-Claude)
• Release Optimization 5.9.1 (#1356) (@​ChrisRackauckas)

Closed issues:

• Manopt linesearch produces NaN/Inf on SymmetricPositiveDefinite manifold (#1135)

Changelog

Sourced from Optimization's changelog.

v4 Breaking changes

1. The main change in this breaking release has been the way mini-batching is handled. The data argument in the solve call and the implicit iteration of that in the callback has been removed, the stochastic solvers (Optimisers.jl and Sophia) now handle it explicitly. You would now pass in a DataLoader to OptimizationProblem as the second argument to the objective etc (p) if you want to do minibatching, else for full batch just pass in the full data.

2. The support for extra returns from objective function has been removed. Now the objective should only return a scalar loss value, hence callback doesn't take extra arguments other than the state and loss value.

Commits

• 9246e25 Release Optimization 5.9.1 (#1356)
• 31a02bd Route solve through the problem's problem_type when wrapping solutions (#1354)
• 7c7e8d3 ConvexOptimization: disciplined parametrized programming (DPP) (#1351)
• 6b1a810 Add AirspeedVelocity benchmarking CI (#1350)
• 800783e ConvexOptimization: lower l1/linf norm and exp/log objectives (#1348)
• 35118bd ConvexOptimization: lower Euclidean-norm objectives through their epigraph (#...
• f09691a Add ConvexOptimization subpackage (disciplined convex programming backend) (#...
• bc4fdb0 Add 32-bit Core CI lane via test_groups.toml (#1346)
• d8b06a5 Release OptimizationManopt 1.6.1 (#1345)
• 33f0d64 Fix OptimizationManopt reporting MaxIters even when converged (#1258)
• Additional commits viewable in compare view