atrfu¶
atrfu solves grey-box optimisation problems with A-TRFu, an advanced trust-region funnel method. Part of the model is written in Pyomo (the glass box). The rest comes from black boxes, such as simulations or legacy code, that can only be evaluated, with no derivatives.
from atrfu import TrustRegionSolver
result = TrustRegionSolver().solve(model, efmap)
The solver replaces each black box with a local linear surrogate built from finite differences. It then solves a sequence of trust-region subproblems with IPOPT and accepts or rejects each step with a funnel. With several black boxes, it measures the error of each one separately and adapts the sampling region from the hypervolume of their errors. A feasibility-restoration phase recovers from runs of rejected steps.
Using atrfu