Manopt.jl Roadmap to 1.0

[kellertuer]

I feel that by now the overall architecture of Manopt.jl seems reasonably stable. In this issue I would like to keep track of (probably breaking) changes and improvements that would need to be done before we release a 1.0. I would currently assume there is at least one, if not two breaking releases before we go for 1.0.
This is also not meant as something that should be rushed – since any breaking change has effects on the eco system.

Feel free to recommend further points as comments below to be added to this list.

• set_parameter was designed a bit ad-hoc and should have a more consistent set of symbols to address semantic an not syntactic elements of sub-problems and sub-states. E.g. :p should become :Iterate.
• get_parameter should be used for accessing sub-solver states besides the result.
• Integration into Optimization.jl (Add Manopt.jl wrapper SciML/Optimization.jl#712)
• All functions from src/functions should either be moved to ManifoldDiff (basic functions) or ManoptExamples (TV, Bézier) (Move Examples. #331).
• Unify passing of optional parts of the objective in high-level solver interfaces #381
• Consider reducing dependencies here by writing an extra package for visualisation, cf Manifold[s]Makie.jl #575 .
• Consider using AlgorithmsInterface.jl before going 1.0 (probably not, that is too much in an early stage – and would be a really major rework here).

Feel free to open more issues of what you would like to see as features within Manopt.jl 1.0

[oscardssmith]

Can I request some help on the Optimization.jl size with SciML/Optimization.jl#906 to get updated to Manopt@0.5?

[kellertuer]

Hi,

I can try to find some time, but for now my experience in getting help there was not the best of experiences. While the support is approximately half there, Manopt is also still not mentioned in the Readme, so for the last few months I did not follow up too much on that.