Module optimization_engine::continuation

Continuation/Homotopy methods for parametric optimization problems

Consider a parametric optimization problem of the general form

Minimize_u f(u; p)
subject to: u in U(p)

Suppose that for a particular value of p, this problem is ill conditioned.

It then makes sense to solve the problem for a sequence of values p_k, which converge to the target value p, starting from an initial value p_0 and using each solution as a warm start for the next optimization problem.

The target value of p can be equal to infinity. In that case we have the so-called "penalty method".

The above parametric problem is accompanied by a function c(u; p), used as a termination criterion. In particular, when c(u; p) drops below a desired tolerance, then the iterative procedure is terminated.

Structs

HomotopyOptimizer	Homotopy optimizer
HomotopyProblem	Homotopy problem definition
HomotopySolverStatus	Solver status of the homotopy method

Enums

ContinuationMode

Continuation mode for free parameters