Module metaheuristics 

General-dimension metaheuristic optimizers (hill-climbing / simulated annealing / Artificial Bee Colony) — global search beyond the fixed-[f64;4] solvers below. Heap-using batch-analytics layer; the engine ontology alignment consumes it. General-dimension metaheuristic optimizers (CI-SKM ch 4/6) — global / non-convex search beyond the fixed-[f64;4] local solvers in this category. Generic local search + simulated annealing work over any state (continuous vectors or combinatorial structures via a neighbour closure — the engine ontology alignment consumes), plus a continuous population optimizer (Artificial Bee Colony).

All minimize the objective. Kernel-class Divergent (branch-heavy search) with the CPU path always present (§13). Deterministic given the seed.

Structs

Rng

Deterministic RNG (LCG + Box-Muller) so searches are reproducible.

Functions

artificial_bee_colony

Artificial Bee Colony (continuous): a swarm explores a box [lower, upper]^d, exploiting good food sources and scouting new ones. Minimizes objective.

hill_climbing

Generic hill-climbing: from initial, repeatedly move to the best improving neighbour until none improves (a local minimum) or max_iter is reached. neighbors enumerates candidate moves; objective is minimized.

simulated_annealing

Generic simulated annealing: accept worsening moves with probability exp(−Δ/T), cooling T ← cooling·T each step, to escape local minima. neighbor proposes a single random move.