Mathieu Besançon

Results 257 comments of Mathieu Besançon

For a more readable format, the polyhedron was created from the following JuMP model: ```julia cons[1] : 26 α[2] + 50 α[5] + 35 α[6] + 25 α[8] + 23...

@sebastiendesignolle I think this was with an old build system. With the new one, you should override the artifact location for cddlib_jll: https://pkgdocs.julialang.org/v1/artifacts/#Overriding-artifact-locations

The UUID to override is here: https://github.com/JuliaBinaryWrappers/cddlib_jll.jl/blob/main/Artifacts.toml So this one: d404927e6c241215ee2871ba75c33e56548a63f0 I think

in a file `.julia/artifacts/Overrides.toml`

for the long term, we could transfer the issue to the lib itself to make this a runtime parameter?

In some cases (if you know bounds on your primal and dual expression), rewriting using big Ms & binaries can be a decent option, for which having the two expressions...

Considered handled by #67 or do we leave this open?

This is likely not gonna be AD-friendly, which will be problematic for projection_on_set more than projection_gradient_on_set

I think somethnig close to what you have in there, with an additional argument for the distance and we are good, so ```julia projection_on_set!(y, ::DefaultDistance, v::AbstractVector{T}, ::MOI.Zeros) ```

has been replaced by https://github.com/ZIB-IOL/FrankWolfe.jl/pull/363 and the recent #415