Repository navigation
Hypatia v0.11.1
Merged pull requests:
- Support Julia builds without GPL libs (
USE_GPL_LIBS=0) (#882) (@oscardssmith) - Add Dependabot to GitHub actions (#883) (@odow)
- Add a CI job for USE_GPL_LIBS=0 (#884) (@odow)
- bump compat of DocStringExtensions (#885) (@araujoms)
- Fix tall NormSpectral/NormNuclear constraints and MOI.modify on Hermitian PSD (#886) (@chriscoey)
- Apply the sparse QR row permutation to c exactly once (#887) (@chriscoey)
- Fix WSOSInterpEpiNormOne crash when a Hessian block's Cholesky fails (#888) (@chriscoey)
- Replace the MOI wrapper's set transformations with a bridge (#889) (@chriscoey)
- Run the cone barrier tests, and split CI into parallel jobs (#890) (@chriscoey)
- Fix divided differences for small or partly clustered eigenvalues (#891) (@chriscoey)
- Speed up WSOSInterpEpiNormOne's inverse Hessian product (#894) (@chriscoey)
- Compute Δ2! without a cutoff and retune the Δ3! and Δ4_ij! cutoffs (#895) (@chriscoey)
- Evaluate the Chebyshev basis in reference coordinates on sampled domains (#896) (@chriscoey)
- Use CHOLMOD's permuted factor in sparse LinMatrixIneq oracles (#897) (@chriscoey)
- Reduce cancellation in the power cones' inverse Hessians and normalize their powers (#898) (@chriscoey)
- Assert the concrete cache types in PosSemidefTriSparse and EpiPerSepSpectral oracles (#899) (@chriscoey)
- Scale the off-diagonals in the sparselmi example's sparse PSD constraints (#900) (@chriscoey)
- avoid Hermitian matrix multiplication in LMI cone (#902) (@araujoms)
- Take the Hermitian part of scaled blocks in PSD-type cone oracles (#903) (@chriscoey)
- Compute MatrixEpiPerSquare's oracles from the Cholesky factor of Z (#904) (@chriscoey)
- Compute the PSD-type inv_hess_prod! through the Cholesky factor (#905) (@chriscoey)
- Batch WSOSInterpNonnegative's rank-one Hessian terms (#906) (@chriscoey)
- Bound the proximity before factorizing the Hessian in the step search (#907) (@chriscoey)
- Map observations to reference coordinates in the densityest example (#908) (@chriscoey)
- Clarify the inv_hess_prod! accuracy comments from #905 (#909) (@chriscoey)
- Solve before forming the product in the CHOLMOD sparse PSD dder3 (#910) (@chriscoey)
- Prep for v0.11.1 (#911) (@araujoms)