Docstrings

TOBSAlg()

Topology optimization of binary structures (TOBS), a sequential-linearization heuristic for binary nonlinear programs. Each iteration linearizes the objective and the inequality constraints around the current design and solves the resulting binary linear program (a MILP, via TOBSOptions' optimizer) under a move limit; infeasible subproblems trigger a feasibility-restoration phase. All variables must have bounds 0 and 1. Equality constraints are not supported.

TOBSOptions(; kwargs...)

Options of the TOBSAlg heuristic:

• movelimit: maximum move limit per iteration as a fraction of the number of variables (at most movelimit * n variables may flip). Default 0.1.
• pastN: number of past iterations used for the moving average of the objective history. Default 20.
• convParam: convergence tolerance on the moving average of the relative objective change over the last pastN iterations. Default 0.001.
• constrRelax: constraint relaxation applied to the linearized constraints (a violated constraint must remove a fraction 1 - constrRelax of its violation each iteration). Default 0.9.
• timeLimit: time limit (seconds) of each MILP solve. Default 1.0.
• optimizer: the JuMP optimizer type used for the MILP subproblems. Default HiGHS.Optimizer.
• maxiter: maximum number of outer iterations. Default 200.
• timeStable: average the objective gradient with the previous iteration's gradient (time-stability filter) when true. Default true.
• relaxDecay: geometric factor shrinking an inflated movelimit/ constrRelax back toward their defaults after successful steps. Default 0.9.
• maxInfeasible: consecutive infeasible subproblems before switching to the feasibility-restoration phase. Default 5.