Christian Schilling
Christian Schilling
We probably do not want to take the last commit. Here are the results (random numbers, hence not really comparable; but note the higher allocations): ```julia using LazySets, BenchmarkTools P...
The code I added in this PR did *not* use keyword arguments, and hence it was fast. I added a commit that uses them now, but I do not want...
> This branch is superseded by #1986. @schillic shall we close it? #1986 just removes the `atan` implementation. This branch contains new features.
It might be useful to first add #1318. Otherwise we would fall back to the default plot recipe for `LazySet` which would plot the convex hull.
Unions became `LazySet`s in #3022. There is still some useful code here, so I keep the branch.
I created #2707 for the concrete operation. So this issue considers the lazy operation only.
> Isn't this a problem with dispatch? Why? > Perhaps the "fix" is to change the default to be `
As commented [here](https://github.com/sisl/NeuralVerification.jl/pull/190#issuecomment-837180086) this is a known behavior which is suboptimal, but it will still take a while until `Complement
This works now. ```julia julia> HS = HalfSpace([-1.0, -1.0], 0.0) HalfSpace{Float64, Vector{Float64}}([-1.0, -1.0], 0.0) julia> Complement(HS) == Complement(HS) true ```
If manual axis limits are specified, sets in a flowpipe are sometimes not plotted. (The exact condition is not clear to me. Maybe if they do not fit? But plotting...