Upgrade to ProximalAlgorithms v0.5

[dhanak]

• drop support for ForwardBackward optimizer;
• add support for PANOCplus;
• make PANOCplus the default optimizer;
• stabilize unstable unit test.

[codecov]

Codecov Report

Merging #40 (e2e8a00) into master (d3a2f64) will increase coverage by 0.37%.
The diff coverage is 92.00%.

❗ Current head e2e8a00 differs from pull request most recent head 50f844b. Consider uploading reports for the commit 50f844b to get more accurate results

@@            Coverage Diff             @@
##           master      #40      +/-   ##
==========================================
+ Coverage   91.87%   92.24%   +0.37%     
==========================================
  Files          16       17       +1     
  Lines         480      503      +23     
==========================================
+ Hits          441      464      +23     
  Misses         39       39              

Impacted Files	Coverage Δ
src/StructuredOptimization.jl	100.00% <ø> (ø)
src/arraypartition.jl	62.50% <ø> (ø)
src/syntax/terms/proximalOperators_bind.jl	92.42% <88.00%> (+0.11%)	⬆️
src/solvers/minimize.jl	94.73% <92.30%> (ø)
src/calculus/precomposeNonlinear.jl	88.88% <100.00%> (+0.42%)	⬆️
src/solvers/build_solve.jl	95.45% <100.00%> (ø)
src/syntax/terms/term.jl	77.77% <100.00%> (ø)

... and 5 files with indirect coverage changes

[lostella]

Thanks @dhanak! I guess we can bump the dependency on AbstractOperators too.

I’m not entirely sure why we need to drop support for ForwardBackward?

[lostella]

Suggested change

	AbstractOperators = "0.1 - 0.2"
	AbstractOperators = "0.3"

[dhanak]

Sure. In fact, I requested a version bump on AbstractOperators, I'm glad it went through so quickly.

[dhanak]

I’m not entirely sure why we need to drop support for ForwardBackward?

ProximalAlgorithms dropped support for an external A matrix, ForwardBackward is now just f(x) + g(x), so the transformations in StructuredOptimzation don't work with FB anymore. There are the DouglasRatchford and the DRLS solvers, which could be added to the suite, though.

[dhanak]

I’m not entirely sure why we need to drop support for ForwardBackward?

ProximalAlgorithms dropped support for an external A matrix, ForwardBackward is now just f(x) + g(x), so the transformations in StructuredOptimzation don't work with FB anymore. There are the DouglasRatchford and the DRLS solvers, which could be added to the suite, though.

[dhanak]

Please note that I won't be available in the next 10 days or so.

[lostella]

Thanks @dhanak!