JuliaSmoothOptimizers / JuliaSmoothOptimizers/ADNLPModels.jl
Use better colorings
- Dominant language
- Julia
- Stars
- 44
- Forks
- 19
- PR merge metrics
- No merged PRs in 30d
Description
As of #244, column coloring is used everywhere, for Jacobians and Hessians alike. However:
- For Jacobians, row coloring + reverse mode may be faster than column coloring + forward mode, depending on the input and output dimensions
- For Hessians, star coloring can leverage the symmetry of the matrix to use fewer colors, but it requires a more clever decompression algorithm
SparseMatrixColorings.jl also provides [decompression utilities](https://gdalle.github.io/SparseMatrixColorings.jl/stable/api/#Decompression), which you may want to use for column and star coloring instead of recoding them.
Of course my recommendation is to use DifferentiationInterface.jl and stop worrying about all of that
Contributor guide
Assessment
This issue has not been assessed yet.