Open anriseth opened 7 years ago
@anriseth I may give this a try. Did you have particular motivation for this algorithm vs others? Would this be a priority to implement?
Ah, great. Thanks for the continued improvements :)
The motivation for this algorithm was that it promised to better populate the Pareto surface in higher dimensions than what NBI could. If I remember correctly it is also better at adapting to situations where the Pareto front is nonconvex or disconnected.
It is difficult to know whether it is worth the effort, however. I have not encountered applications where the currently implemented methods are not sufficient.
Depending on your and other users' usecases, it may be more valuable to facilitate for a larger type of optimization problems and algorithms, see, for example, https://github.com/anriseth/MultiJuMP.jl/issues/1
ok interesting, I have to dig more into the paper
Moved from README.md.
References: