Juniper.jl
Juniper.jl copied to clipboard
New primal heuristic: Feasible Rounding Approach by Shrink-Opimtize-Round
A new primal heuristic was recently proposed in The granularity concept in mixed-integer optimization. The idea is to define a subset of the feasible set so it is are guaranteed that at-least one rounding exists that is in the feasible set, optimize this relaxation and then round once to get to a feasible solution faster. This is trivial for Linear constraints but doing it for non-linear constraints might be harder or require an iterative approach, (i haven't done the math). I will also suggest this for Alpine.
Great suggestion, would love to get it added at some point.